+ All Categories
Home > Documents > Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării...

Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării...

Date post: 04-Jan-2020
Category:
Upload: others
View: 2 times
Download: 0 times
Share this document with a friend
114
1 HABILITATION THESIS SERVOSYSTEMS FOR MOTION CONTROL IN THE ROBOTS’ TECHNIQUE Domain: Electrical engineering Author: Prof.dr.ing. Aurel Fratu University Transilvania of Brasov BRASOV, 2015 Universitatea Transilvania din Braşov
Transcript
Page 1: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

1

HABILITATION THESIS

SERVOSYSTEMS FOR MOTION CONTROL IN THE ROBOTS’

TECHNIQUE

Domain: Electrical engineering

Author: Prof.dr.ing. Aurel Fratu

University Transilvania of Brasov

BRASOV, 2015

Universitatea Transilvania din Braşov

Page 2: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

2

CONTENT

(A) Rezumat 3

(B) Scientific and professional achievements and the evolution and development

plans for career development

7

(B-i) Scientific and professional achievements 7

Introduction 7

Chapter 1 Dynamic models for control system design of integrated robot and

drive systems

1.1 Dynamics of the nonlinear robotic systems with constraints

1.2 Design of flexible drive systems for feed-forward control: a

mechatronic approach

14

22

Chapter 2 Haptic interfaces

2.1 Haptic system modeling and stability analyze

2.2 Hybrid force/position control of the haptic scissors-robot

34

44

Chapter 3 The reciprocal robots collision avoidance……...........................................

3.1 Collision-avoidance using redundant robots

3.2 On-line collision avoidance using Programming by Demonstration

62

73

Chapter 4 Modeling and robots behavioral simulation in visual

environments ...............................................................................................

4.1 System concept development with virtual prototyping

4.2 Virtual Prototyping for robotic systems

83

91

(B-ii) The evolution and development plans for career development.......................

99

(B-iii) Bibliography

103

Page 3: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

3

(A) Rezumat

În cadrul tezei actuale de abilitare, realizările mele stiintifice, profesionale şi academice sunt

cuprinse într-o perioadă, începând cu anul 1994 (data la care teza de doctorat a fost susţinută,

apoi atestată prin Ordinul ministrului 6082 / 21.04.1994) până în prezent (2014). Unele dintre

granturile mele naţionale şi internaţionale de cercetare, articole importante, cărţi, brevete, cursuri

predate sunt, de asemenea, detaliate în contextul stadiului actual al domeniului stiinţific al

ingineriei electrice, cu accent pe aspectele inovatoare şi contribuţii personale.

Prima parte (A) a tezei de abilitare este constituită din prezentul rezumat în limba română.

Cea de a doua parte (B) a tezei curente se referă la realizări stiinţifice, profesionale şi planul de

evolutie şi dezvoltare a carierei şi este formată din trei secţiuni (B-i), (B-ii) si (B-iii).

Sectiunea (B-i) conţine patru capitole care tratează patru teme principale de cercetare, prefatate

de o introducere.

In introducere se prezintă o privire de ansamblu asupra activităţii autorului tezei de abilitare, cu

referire la cele mai importante direcţii de cercetare - clasificate în patru teme principale -

realizările profesionale şi academice, disciplinele nou introduse, cursuri predate, contribuţia la

dezvoltarea curriculei universitare, misiuni de profesor invitat, studenţi in stagiu, conducerea

lucrărilor de diplomă şi disertaţie, cooperare internaţională, activităţi de management, etc.). Cele

mai importante aspecte mentionate sunt: un număr de 73 de articole de cercetare publicate în

perioada mentionată mai sus, 4 granturi de cercetare (2 ca director de contract) si 8 cărţi.

In capitolul 1 este prezentată urmatoarea temă:

Modele dinamice pentru proiectarea sistemului de control al sistemelor integrate de

roboti şi sisteme de acţionare

Principala sarcină a sistemului de control a mişcării robotului manipulator este de a genera

mişscare în spaţtiul sarcinii robotice, cu o comandă dată la nivelul sistemelor de acţionare.

Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda

liniarizării folosind feedback-ul. Ideea de bază a acestei tehnici este aceea de a proiecta o lege de

control care anulează /compensează neliniaritătile sistemului şi rezultă un sistem cu circuit

închis cu o dinamică liniară.

In primul subcapitol al capitolului 1 se studiază o ecuație universală de mişcare pentru sisteme

robotizate cu constrângeri. Aici, problema controlului sistemelor neliniare este analizată făra a

lua in calcul tipul de controler care urmează să fie utilizat.

Rezultatele prezentate în această lucrare oferă metode noi şi explicite pentru controlul sistemelor

puternic neliniare.

Al doilea subcapitol al capitolului 1 se referă la proiectarea sistemului mecatronic pentru

controlul mişcării de mare viteză, folosind o buclă feed-forward.

În sistemele integrate de roboti şi sisteme de acţionare principalele caracteristici, în mod special

în controlul vitezei de mişcare, sunt complexitatea dinamicii şi incertitudinile, atât parametrice

cât şi dinamice. Incertitudinile parametrice apar din cunoaşterea imprecisă a parametrilor

cinematici şi a parametrilor de inerţie, în timp ce incertitudinile dinamice apar din cauza

Page 4: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

4

flexibilităţii legăturilor dintre articulaţii, dinamicii actuatorilor, frecării, zgomotului senzorilor şi

necunoaşterii dinamicii mediului.

Abordarea propusă include proiectarea sistemului de acţionare şi control a mişcării. Scopul este

de a proiecta atât transmisia mecanică cât şi motorul electric pentru a atinge specificaţiile de

performanţă cu ajutorul controlerului. Modelul elasto-dinamic invers al sistemului este utilizat

pentru a genera comandă şi pentru a stabili constrângerile de proiectare datorită limitelor fizice.

Proiectarea este realizată folosind frontiera Pareto care permite o alegere optimă a motorului,

precum şi o dimensionare optimă a mecanismului de transmisie. In exemplul prezentat se arată

că această abordare de proiectare conduce la obţinerea unui sistem robot integrat robust, tolerant

la erori de modelare şi identificare.

In capitolul 2 este prezentată următoarea temă:

Interfeţe haptice.

In primul subcapitol al capitolului 2 se abordează problema de stabilitate a interfeţei haptice în

contact cu un perete virtual. Această lucrare prezintă o imagine de ansamblu cu privire la

metodologie şi la regiunea de stabilitate afectată de anumiţi parametri: coeficientul de

amortizare fizică, constanta de întârziere, caracteristici dinamice ale operatorului uman.

Scopul acestei lucrări este de a analiza influenţa parametrilor dinamici asupra granitelor de

stabilitate. Acest studiu arată modul în care factorii cum ar fi coeficientul de amortizare fizică,

temporizarea şi parametrii operatorului uman afectează stabilitatea unui sistem haptic. Influenţa

tipului de vibraţii asupra stabilităţii, este de asemenea studiat în această lucrare. Rezultatele sunt

efectuate asupra parametrilor fizici ai unei interfete haptice de tip PHANTOM®.

În al doilea subcapitol al capitolului 2 este descrisă o metodă pentru ilustrarea senzaţiei tactile a

operatiei de tăiere, folosind un tip de foarfece-robot virtual. Este prezentat un nou algoritm

pentru a fi utilizat în manipulări generale interactive ale obiectelor deformabile. Algoritmul a

fost dezvoltat pentru a afişa deplasarea simultană de translaţie (de înaintare) şi de rotaţie (de

decupare) pentru o simulare realistă a operaţiei de tăiere. Am arătat că animaţiile interactive pot

fi utilizate pentru a simula funcţionarea dispozitivelor de tip foarfece, folosind reacţia de forţă /

cuplu. Prin această lucrare se obţine experienţă in procesul de proiectare în mediul virtual şi se

determină noi caracteristici ale posibilelor dispozitive de tip foarfece-robot care folosesc reacţia

de cuplu.

In capitolul 3 este prezentată urmatoarea temă:

Evitarea reciprocă a coliziunii robotilor

Scopul primului subcapitol al capitolului 3 este de a analiza problema cinematicii inverse a

roboţilor redundanţi şi de a defini sub-sarcina robotică de-a evita obstacolele, folosind mişcarea

spontană. Strategia propusă permite utilizarea gradelor de libertate redundante, astfel încât un

manipulator să poată evita obstacolele în timp ce efectorii finali urmaresc traiectoria dorită. Se

presupune că obstacolele, aflate în spaţiul de lucru al manipulatorului sunt statice. Această

strategie este aplicată şi atunci când configuraţia manipulatorului poate fi influenţată de cerinţe

suplimentare, cum ar fi limitările la nivelul articulaţiilor, etc.

Eficacitatea strategiei propuse este analizată din punct de vedere teoretic şi ilustrată prin

simularea mişcării - intre obstacole simetrice - a unui manipulator redundant cu patru grade de

Page 5: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

5

libertate. Se constată că strategia propusă - în scopul evitării coliziunii în timp ce efectorii finali

urmăresc traiectoria impusă - este eficientă şi practică.

Cea de a doua lucrare dezvoltă o metoda originala de-a evita coliziunea, bazata pe conceptul de

imitare a miscarii. Operatia de imitare a miscarii, presupune existenta unui demonstrator si a

unui imitator. În demersul nostru demonstratorul este prototipul robot virtual si imitatorul este

robotul fizic real. Intrările pentru procedura de imitatie sunt traiectorile virtuale articulare,

reprezentate ca o secvență de valori unghiulare ale articulatiilor sale. Imitatorul imită miscarea

originală capturata de la prototipul virtual, si în acelasi timp, respectă constrangerile fizice.

Punctul forte al metodei propuse este că oferă o re-planificare on-line a miscării în fata

perturbatiilor (obstacolelor) spatio-temporale, folosind anumite scenarii. Miscarea modelului

robot virtual trebuie să acopere toate situatiile posibile - inclusiv prezența obstacolelor

accidentale – pentru care robotul fizic va trebui să genereze miscari similare.

Metoda de evitare a coliziunilor, bazată pe imitarea de prototipuri virtuale, trebuie încă

îmbunătățită pentru a răspunde la întrebarea: cum pot fi actualizate on-line modele sistemelor

dinamice virtuale pentru a lua în considerare evenimentele dinamice din mediul real ? Autorul

estimeaza ca programarea prin imitatie, complet automatizata, folosind un sistem suficient de

robust pentru aplicații practice, nu va deveni realitate inainte de sfârsitul acestui deceniu.

In capitolul 4 este prezentată urmatoarea tema:

Modelarea si simularea comportamentala a robotilor în medii vizuale.

Scopul primului subcapitol al capitolului 4 este de a evalua problema prototiparii virtuale.

Prototiparea virtuala reprezinta un mijloc pentru a descrie calitativ comportamentul produsului

sub diverse aspecte.Prototiparea virtuala este o disciplină de inginerie software, care implică

modelarea unui sistem, simularea și vizualizarea comportamentului său în conditii de

funcționare reale si rafinarea proiectului printr-un proces iterativ. Comportamentul full-motion

a sistemelor robotice complexe poate fi analizat înainte de a construi un prototip hardware real.

Utilizatorii pot explora rapid multe variante de proiect, de testare si rafinare până cand sistemul

robotic este optimizat. Acest lucru poate ajuta la reducerea timpului si a costurilor de dezvoltare

a unui nou produs.

Scopul propus in cel de-al doilea subcapitol al capitolului 1 este de a sublinia rolul simulării în

diferite domenii ale roboticii. Avand capacitatea de a simula, se deschide o gamă largă de

optiuni pentru rezolvarea multor probleme in mod creativ. Se poate modela, vizualiza, investiga

si testa un braț de robot, chiar dacă aceasta nu există. Se pot vedea performanțele unui sistem

robotic înainte de a fi construit. In mediul virtual este posibil ca soluiile alese sa esueze sau

sistemul proiectat sa „‟explodeze‟‟ fara a avea consecinte asupra prototipului virtual.

Simularea a fost recunoscută ca un instrument important în robotica, in proiectarea de noi roboti,

investigarea performantele lor si în proiectarea aplicațiilor acestora. Simularea permite studierea

caracteristicilor structurale si a functiilor sistemelor robotice, la diferite niveluri de detaliu,

fiecare situatie avand cerințe diferite pentru instrumentele de simulare. De exemplu, un proces

rapid - cum ar fi deplasarea unei articulatii robot - poate fi încetinit pentru a observa toate

detaliile in "slow motion".

Page 6: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

6

Pe masura ce complexitatea sistemului creste, rolul simularii devine tot mai important si

folosind instrumentele de simulare se pot îmbunătăți proiectele.

În al doilea subcapitol al capitolului 4 se propune folosirea mediului vizual de programare

Delphi, pentru a modela si simula comportamentul bratelor robotice de tip serial si de tip paralel.

Simulările realizate pentru aceaste aplicatii au servit la dezvoltarea metodelor sistematice pentru

crearea de noi sisteme robotizate.

Rezultatele acestei analize constă în identificarea proprietătilor si a constrângerilor care pot

aparea in cazul prototipurilor fizice homonime. Structurile virtuale create în această lucrare

includ aplicatiile cu brate robotice de tip serial ( exemplificare - pentru statii de triere bagaje) si

respectiv cu brate robot paralele ( exemplificare - pentru platforme de tip hexapod).

Este important de retinut ca principalele rezultate obtinute în domeniul primei directii de

cercetare Modele dinamice pentru proiectarea sistemului integrat de control robot si sisteme de

acționare sunt identificate în sectiunea (B-iii), de existenta cuvintelor control si robot în titlul

unui număr de 19 articole publicate de autorul tezei de abilitare. De asemenea, un contract de

cercetare abordeaza o temă similară.

A doua direcție de cercetare, si anume Interfetele haptice a fost tratata în 5 articole publicate de

autorul tezei de abilitare (identificate de existența cuvântului haptic în titlu lor, în sectiunea (B-

iii).

Cea de a treia problematica, aceea de Evitare reciproca a coliziunii robotilor a fost investigata

într-un număr de 8 articole publicate de autorul tezei de abilitare (a se vedea referintele cu

existenta cuvintelor evitare coliziuni în titlu lor, în sectiunea (B-iii).

Rezultatele privind ultima directie de cercetare, Modelarea si simularea comportamentului

robotilor în medii vizuale, au fost publicate de autorul tezei de abilitare în 6 articole (a se vedea

referintele cu existența cuvintelor modelare si simulare în titlul lor, în sectiunea (B-iii).

Cea de-a doua sectiune (B-ii) prezinta Planul de evolutie si dezvoltare a carierei

Pentru viitor cercetarea unor noi strategii de comunicare - între mediul informatic virtual si cel

fizic real - sunt avute in vedere pentru a fi profund studiate (cu aplicații pentru interfete haptice

si aplicatii care vizeaza evitarea reciproca a coliziunilor). Noi solutii de implementare vor fi

propuse (de exemplu platforma de programare prin imitatie, bazată pe o idee originală

brevetata). De asemenea, am de gând să investigheze noi domenii de cercetare în colaborare cu

cercetatori de la Universitatea din Valenciennes si Universitatea din Reims - Franta.

Cea de-a treia sectiune (B-iii) este dedicată referințelor bibliografice.

Page 7: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

7

(B) Scientific and professional achievements and the evolution and development plans for

career development

(B-i) Scientific and professional achievements

Introduction

The interest in creating such an overview is mixed. By selecting the most interesting approaches,

we want to focus the attention to new techniques and methodologies that may be of high interest

to the researchers in the field of collision avoidance. Moreover, this selection can be a useful

indicator of the areas that will constitute the future research trends.

The title of my PhD thesis was Reaserch on positioning system with synchronous servo-motor

and micrtocomputer for industrial robots. I presented it publicly at the "Transilvania" University

of Brasov in 1994. Therefore the following overview of my scientific interests and work will be

made starting from the above mentioned date up to present (2015).

During this period, Dynamic models for control system design, Haptic devices, Robots

collision avoidance, Modeling and robots behavioral simulation, were the main research

topics.

They represent as well my teaching activities. For example, during 2009 I have proposed,

elaborate the syllabus and taught two new courses: “Manipulators and Robots Command and

Control” and “Visual programming with Applications in Robotics”, within the framework of the

bachelor program.

The main developed applications of the above mentioned topics are grouped into four categories

that constitute the chapters of the (B-i) section, as follows:

1. Dynamic model for control system design of integrated robot and drive systems

1.1 Dynamics of the nonlinear robotic systems with constraints

1.2 Design of flexible drive systems for feed-forward control: a mechatronic

approach

The contributions to the field of Dynamics of the nonlinear robotic systems with constraints are

presented in subchapter 1.1 of chapter 1 of the current thesis. The first article, presented in

details in [34], deal with the analytical model of motion in terms of second-order differential

equations. This methodology has been inspired by results in analytical dynamics. The explicit

closed-form expression provides the entire set of continuous tracking controllers that can exactly

track a given trajectory description, assuming that the system‟s initial conditions satisfy the

Page 8: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

8

description of the trajectory. The explicit closed-form expressions for the controllers can be

computed in real time.

Closed-form expressions for all the continuous controllers required for trajectory tracking for

nonlinear systems do not make approximations. Furthermore, no approximations or linearization

are made here with respect to the trajectory that is being tracked, which may be described in

terms of nonlinear algebraic equations or nonlinear differential equations.

Equation obtained is the general, explicit equation of motion for constrained robotic systems

with non-ideal constraints. Here, we do not have any restriction regarding the positive

definiteness of the symmetric mass matrix. One can obtain the following results:

First result The general equation of motion of a constrained robotic mechanical system, whether

or not the mass matrix that arises in the description of the unconstrained motion of the system is

singular.

Second result When the mass matrix has full rank, the equation of motion of the constrained

system is unique.

Moreover, the approach arrives not just at one nonlinear controller for controlling a given

nonlinear system, but also at the entire set of continuous controllers that would cause a given set

of trajectory descriptions to be exactly satisfied.

The contributions to the field of Design of flexible drive systems for feed-forward control are

presented in subchapter 1.2 of chapter 1 of the present thesis. The article [56] published in a

Proceeding conference was chosen as the illustrative paper for this topic. It achieved a number

of 4 citations in Thomson Web of Knowledge database.

This approach led us to have an over-sizing or performances lower than specifications. The

mechanism is indeed sized to transmit a mechanical power (force, velocity). On top of that, to

avoid any interaction with the control (resonance excitation, spill-over phenomenon the

mechanism can be over-sizing to place the resonant frequencies high enough, what leads to have

a high mechanism inertia and therefore an oversized motor. In the case of the mechanism is not

oversized (weaker inertia, lower resonance frequencies), a control law damping the vibratory

modes will be able to be used.

A state space feedback control of flexible robot will prove to be often delicate to implement,

what restricts its use outside laboratories and space applications. An attractive approach in set

point tracking is the use of a feed-forward term established from dynamic inverse model as it

was proposed for the elastic servomechanisms and the flexible link robots. In addition, the

authors of this article offer a modification of the dynamics of the arm of robot to be able to

generate a simpler feed-forward control. Mechanical modification is obtained by addition of

punctual masses along the arm in order to achieve a relation specified between the frequencies

of resonance. These works show interest to shape the system dynamics to achieve better results

with control.

A mechatronic design method of flexible drive systems dedicated to feed-forward control has

been proposed. The originality of the approach is to enter at the same time the question of design

and the control to reach performance requirements. This approach uses the inverse elastic

dynamic model of the drive system and its physical limits (motor saturations, mechanical

resonance, etc.). Based on this formulation, a two-step design method has been established and

applied to an example in order to show the characteristics (performances and robustness) of a

design solution.

Page 9: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

9

One of the extensions of this approach of total optimization concerns nonlinear mechanisms and

arms of flexible robot with, in this case, a particular attention to non minimal-phase behavior.

Other attractive subject undertake within this framework (Dynamic models for control system)

was related to the development of an integrated environment for the visually control movement

of the robotic systems. The results were published in several articles e.g. [70] or [71], the former

published in an ISI Proceeding, achieving citations in Thomson Web of Knowledge database.

It is merit noticing that the main results achieved in Control system design of integrated robot

and drive systems field are identified in the section (B-iii), by the existence of robot control

words in the title and are in number of 19 papers published by the author of this thesis.

The research was focused on finding robust solutions for the following subsystems: Motion

control of the synchronous motor and adequate inverter demanded [Brevet RO112454/

29.08.1997].

The international cooperation regarding this research direction was done with prestigious

universities and research institutes. Among them:

- LAMIH (Laboratory of Industrial and Human Automation control, Mechanical

engineering and Computer Science) from University of Valenciennes, France;

- Centre in Information and Communication Science and Technologies (CRESTIC)

Université de Reims Champagne-Ardenne , France

2 Haptic system

2.1 Haptic system modeling and stability analyze

2.2 Hybrid force/position control of the haptic scissors-robot

The contributions to the field of Haptic system modeling and stability analyze are presented in

subchapter 2.1 of chapter 2 of the present thesis. The article [62] published in Proceeding

Conference is chosen as the first illustrative paper for this topic. It achieved a number of 4

citations in Thomson Web of Knowledge database.

In this approach, the virtual environment is as a part of a model which consider the virtual

impedance as a linear spring and damper. The stability region is represented by the stability

boundaries in the stiffness-damping plan of the virtual impedance.

Hereby, the influence of parameters such as physical damping and time delay on stability

boundaries is evaluated with the transfer function model. The human arm impedance can be

included in the model to study the human dynamic effects. There are a lot of works concerning

the interaction of the human operator, the haptic device and the virtual environment.

The existence of internal vibration of haptic device has been studied in recent years, but these

researches have not clearly explained the influence of the vibration mode parameters. The goal

of this paper is to analyze the influence of the dynamic parameters on the stability boundaries.

This study shows how the factors such as physical damping coefficient, time delay and human

operator parameters affect the stability of a haptic system. The influence of vibration mode on

Page 10: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

10

the stability is also studied in this paper. The results are carried out on the physical parameters of

a PHANTOM®.

This paper presents an overview on the effects of dynamic factors on haptic interface stability.

The impedance model used to compute the force feedback includes a virtual stiffness and a

virtual damping.

The stability boundaries of both virtual parameters are derived from a sampled data model with

the Routh-Hurwitz criterion in an analytic way or with the gain margin criterion in a numeric

way. The stability analysis is first based on rigid model of device to show the effects of physical

damping, time delay and human operator. The first vibration mode due to elastic transmission of

haptic device has been also studied. The results allow us to conclude that the vibration modes of

device have an intricate effect to the stability of interface. The analysis of the anti-resonant

frequency effect leads to explanations of qualitative modification of stability boundaries.

Page 11: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

11

This article has provided engineers some basic guidelines in choosing design specifications of

haptic system. In future works the effect of others physical parameters (e.g; time delay, human

model) will be studied with a model included vibration modes.

This article was supported in part by the International Campus on Safety and Intermodality in

Transportation, the European Community (through the FEDER European Funds for Regional

Development), the Délégation Régionale à la Recherche et à la Technologie, the Ministère de

l‟Enseignement supérieur et de la Recherche, the région Nord Pas-de-Calais and the Centre

National de la Recherche Scientifique (CNRS). The international cooperation regarding this

research direction was done between Transilvania University of Brasov and University of

Valenciennes, France.

The research direction, namely Haptic system modeling and stability analyze has been treated in

8 publications (referred by haptic word in title, in section (B-iii).

The contributions to the field of Hybrid force/position control of the haptic scissors-robotare

presented in subchapter 2.2 of chapter 2 of the present thesis. The article [30] published in a

Proceeding conference was chosen as the illustrative paper for this topic. It achieved a number of

4 citations in Thomson Web of Knowledge database.

In the first part this section presents an analytical model based on the concepts of contact

mechanics and fracture mechanics to calculate forces applied to scissors during cutting of a plate

of material.

The model considers the process of cutting as a sequence of deformation and fracture phases.

During deformation phases, forces applied to the scissors are calculated from a torque-angle

response model synthesized from measurement data multiplied by a ratio that depends on the

position of the cutting crack edge and the curve of the blades.

In the second part of this section a method for displaying feeling of cutting, using a virtual

scissors type, is described. A new computer haptics algorithm to be used in general interactive

manipulations of deformable objects is presented. The algorithm was developed to

simultaneously display translational and rotational (cutting) displacement for a realistic cutting

simulation. We showed that interactive animations can be used to simulate the functioning of

force/torque - feedback scissors devices.

This paper discusses both the experience with the design process in the virtual environment and

the novel features of the possible portable torque-feedback scissors devices. We will present an

approach for designing a more natural user interface, without resorting to various special haptic

input/output devices.In this example, our scenarios were simple enough so that the absence of

detailed specifications did not obstruct our later prototyping work.

The model was parameterized with data and equations and simulated over time. Simulation

results can be viewed on a graphical monitor, and finally exported to a physical device.

The research direction, namely Hybrid force/position control of the haptic scissors-robot has

been treated in 8 publications (referred by the words haptic control in title, in section (B-iii),).

Page 12: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

12

3 Reciprocal robots collision avoidance

3.1 Collision-avoidance using redundant robots

3.2 Collision avoidance using Programming by Demonstration

The contributions to the field of the Collision-avoidance using redundant robots are shown in

subchapter 3.1 of chapter 3, of the current thesis. The aim of the first presented work, presented

in details in [60], is to identify the strategy that allows the use of redundant degrees of freedom

such that a manipulator can avoid obstacles while tracking the desired end-effectors trajectory. It

is supposed that the obstacles in the workspace of the manipulator are static. The results were

published in an ISI Proceeding, achieving citations in Thomson Web of Knowledge database.

The strategy is based on the redundant inverse kinematics and is discussed by theoretical

considerations. This strategy has the advantage that the configuration of the manipulator can be

influenced by further requirements such as joint limits, etc. In this paper the problem of

redundant inverse kinematics is reviewed and obstacles avoid subtask for exploiting the self

motion are defined.

The contributions to the field of the Collision avoidance using Programming by Demonstration

are presented in subchapter 3.2 of chapter 3 and concern an article published in an academic

journal [69]. In this article, based on original idea, the author proposes a new strategy to robot

collision avoidance using imitation programming paradigm.

To program the desired motion sequence for the physical robot, one captures the motion

reference paths from her virtual robot model and maps these to the joint settings of the physical

robot. Motion imitation requires transfer of a dynamical signature of a movement of the virtual

robot to the physical robot, i.e. the robots should be able to encode and reproduce a particular

path as one with a specific velocity and/or an acceleration profile. Furthermore, the virtual robot

must cover all possible contexts in which the physical robot will need to generate similar

motions in unseen context.

In this paper the graphical simulation prototype system has been developed to support and

demonstrate different aspects of robots behavior when met the obstacles in his work

environment. All of these programs are write in Delphi language and enable the manifestation of

global behaviors.

For itself they include the movement of virtual robot manipulator, the contact with virtual objects

and the collision avoid, into the work environment, with expected or unexpected virtual objects.

The effectiveness of the proposed strategy is discussed by theoretical considerations and

illustrated by simulation of the motion of the four-joint planar manipulators between symmetric

obstacles. It is shown that the proposed collision-free strategy while tracking the end-effector

trajectory is efficient and practical.

A redundancy-resolution scenario was proposed to achieve obstacle avoidance. In the following

simulations the main task consists of tracking the position and orientation trajectories, generated

by linear interpolation, between the initial and final poses. It should be noted that interpolation of

rotations is a much more complex problem than point interpolation. For this reason, we use

Page 13: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

13

simple linear interpolation for both translation and rotation, which nevertheless leads to

satisfactory results.

In the simulation presented in this section the redundancy resolution is implemented in closed-

loop. One solution to this problem is online transmission of a robot configuration to a

workstation running a graphics visualization of the arm how serves as a virtual environment; the

graphics model of the robot mirrors the exact motion of the arm, and the environment can be

modeled in the graphics program.

The problematic, collision avoidance has been investigated in a number of 8 papers (see

references with collision avoidance words in title, from section (B-iii)) and one patent [Method

and installation for programming the motion paths of robotic arm articulations, Patent

Number(s): RO – 129121(A0)]. The novelty of the proposed system in this patent is related to

robot programming by imitation based on her virtual homonym

.

4 Modeling and robots behavioral simulation

4.1 System concept development with virtual prototyping

4.2 Virtual Prototyping for robotic systems

The contributions to the field of System concept development with virtual prototyping are

presented in subchapter 4.1 of chapter 4, of the current thesis.

The aim of this approch, presented in details in [55], and published in an international colloque is

to identify the strategy that allows the rapid prototyping to create a three-dimensional virtual

model.

Through careful development of details, 3D graphics applications can generate virtual prototypes

that provide a useful tool supporting engineering and specialty disciplines analysis for a wide

variety of system developments, both early in the concept development phase and later in

advanced engineering. Virtual prototyping permits users, designers, and logisticians to visualize

two and three dimensional relationships and clearances for joint multi-disciplinary analysis. This

early analysis and identification of design issues offers potential to reduce development costs

since mistakes will be made in the computer rather than on full-scale prototypes. Due to design

maturity, virtual prototypes developed in advanced engineering development will be based on

design analysis and known qualities. In contrast, exceptional care must be taken during the

technology assessment and advanced concept phases, when details are less certain, to conduct

thorough research that produces and supports a credible product based on expert opinion,

technology analysis, and traceable assumptions rather than “fantasy.” The development of

worksheets containing source and rationale information on each virtual component is a

disciplined technique that helps to ensure reality is not misrepresented.

The second detailed approach concerns Virtual Prototyping for robotic systems and is based on

an article [33] published in a Proceedings Conference. This approach is presented in subchapter

4.2 of chapter 4, of the current thesis.

The purpose of the paper is to emphasize the role of the simulation in different fields of robotics.

Also we recommend to researchers to use the general dynamic engines and general visual

programming languages for the simulation and visualization of robotic systems.

This paper discusses the using of the visual programming Delphi environment, for facilitating

intuitive robot programming. From a robotic point of view, the virtual robotic structures created

Page 14: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

14

were simplified and their analyzing allows a better description of the behavior of real robotic

structures.

The simulations prepared for this application were also made for the development of the

systematic methods for creating the new robotic systems. The results of this investigation consist

in an identification of the properties and constraints of the virtual robot prototypes.

The virtual structures created in this paper encompass the behavior of robotic systems. They may

present a behavior which is qualitatively different from the behavior that the robots currently

have.

In the end, the simulation in robotics has reached a very important role and by using different

simulation software, the present and future capabilities of complex robotic systems can be

significantly improved.

In the virtual prototypes gestures imitation represent one of the most attractive technology in the

field. This work proposes a fingertip -based approach for programming the motion paths of

robotic arm articulations. The novelty of the proposed system is related to robot programming

by imitation based on her virtual homonym.

The results regarding the last research direction, Virtual Prototyping for robotic systems for

robotics, were published in 16 scientific publications by author (see references signed by author,

with virtual prototyping words in title, in section (B-iii).

The international cooperation regarding this research direction was done with prestigious

research group; among them GREAH (Groupe de Recherche en Electrotechnique et

Automatique du Havre) de l'Université du Havre, France.

I have a dynamic activity for creative work: I am owner of 22 brevets of invention. For this

activity I was awarded with Elite Inventor Brevet by Commission of Invention of Romanian

Academy.

From 2006 until present (2015), I was visiting professor, in six years for one month.

From 1994 to present I am a member of different academic/professional societies as SIR, SRR,

IPIMEA and ACER.

Page 15: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

15

Chapter 1 Dynamic models for control system design of integrated robot and drive

systems

1.1 Dynamics of the nonlinear robotic systems with constraints

Introduction

Most of the robotic applications are restricted to slow-motion operations without interactions

with the environment. This is mainly due to limited performance of the available controllers in

the market that are based on simplified system models. To increase the operation speed with

more servo accuracy, advanced control strategies are needed.

The main specifically properties in the speed motion control of the robots systems are the

complexity of the dynamics and uncertainties, both parametric and dynamic. Parametric

uncertainties arise from imprecise knowledge of kinematics parameters and inertia parameters,

while dynamic uncertainties arise from joint and link flexibility, actuator dynamics, friction,

sensor noise and unknown environment dynamics.

Robot‟s motion trajectories are typically specified in the task space in the terms of the time

history of the end-effector‟s position, velocities and acceleration. Operational space (also known

as task space) is the space in which high-level motion and force commands are issued and

executed.

The operational-space formulation is therefore particularly useful in the context of motion and

force control systems. On the other hand, in the joint space control methods, is assumed that the

reference trajectory is available in terms of the time history of joints positions and orientations of

robot arm.

The natural strategy to achieve task space control goes through two successive stages:

in the first stage, the robot‟s kinematics in the task space variables is passed into the

kinematics corresponding joint space variables, and then;

in the second stage is designed the control in the joint space.

Because of the complexity of both the kinematics and dynamics of the manipulator and of the

task to be carried out, the motion control problem is generally decomposed into three stages:

motion planning,

trajectory generation,

trajectory tracking.

Page 16: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

16

The main problem of motion robot control is to generate the motion in the task space with a

given command at joints level. Motion control of robot arm accomplishes the following

functions:

to find of the corresponding movements in joints;

to generate of control signals for the actuators to produce the input torques;

to synthesize of programmed paths.

For trajectory tracking, the computed reference trajectory is then presented to the controller,

whose function is to cause the robot to track the given trajectory as closely as possible. For

design of the tracking controller, one assumes that the reference trajectory and path have been

pre-computed.

Control of robot manipulators is naturally achieved in the joint space, since the control input are

joint torques. But, the user specifies a motion in the task space, and thus it is important to extend

the control problem to the task space. This can be achieved by different strategies. The more

natural strategy consists of inverting the kinematics of the manipulator to compute the joint

motion corresponding to the given end-effectors motion.

Thus, the methods used to appointment primarily depend on linearization and/or PID-type

control, and they envisage suppositions on the structure of the control effort.

Control of Nonlinear Dynamical Systems

Most physical robotic systems are inherently nonlinear. Thus, control of nonlinear systems is a

subject of active research and increasing interest. However, most controller design techniques for

nonlinear systems are not systematic and/or relate only to very specific cases. The most general

results available for nonlinear processes relate to scenarios in which:

all uncertainty is parametric with a known functional dependence of the state-space

model with respect to the unknown parameter, and

there is no measurement noise nor disturbances.

Under these admittedly restrictive assumptions the results available are quite general. For each

possible value of the parameter, one needs to know how to design a (non-adaptive) controller

that would stabilize the process if the parameter value was known. Such controller should be able

to guarantee input-to-state stability with respect to an appropriately introduced disturbance. For

each value of the parameter, one needs to know how to design a (non-adaptive) output estimator

for the process that would converge to the process output if the parameter value was known. This

is an insignificant subject when the whole state of the process can be measured, but can still be

challenging for nonlinear systems for which the state cannot be measured.

The main challenge that remains open in the control of nonlinear systems is robustness with

respect to disturbances. Although, the algorithms appear to work well in the presence of

disturbances, few stability results are available.

There are few systematic procedures to design controllers/estimators for nonlinear systems that

are robust with respect to disturbances. In addition, there are also few results to analyze the

closed-loop switched nonlinear systems that arise as one switches among different controllers.

Current systematic approach to design controllers for nonlinear systems is feedback linearization.

The basic idea of this technique is to design a control law that cancels the nonlinearities of the

Page 17: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

17

plant and yields a closed-loop system with linear dynamics. However, the technique is not robust

to disturbances and uncertainties in the robot parameters, can yield to uncontrolled dynamics

called zero dynamics and can only be applied to systems verifying certain vector field relations.

The development of controllers for nonlinear robotic systems has been an area of intense

research. Many controllers that have been developed for trajectory tracking of complex nonlinear

and multi-body systems rely on some approximations and/or linearization. Most control designs

restrict controllers for nonlinear systems to be affine in the control inputs. Often, the system

equations are linearized about the system‟s nominal trajectory and then the linearized equations

are used along with various results from the well-developed theories of linear control. While this

often works well in many situations, there are some situations in which better controllers may be

needed. This is especially so when highly accurate trajectory tracking is required to be done in

real time on systems that are highly nonlinear such the robotic systems.

In the robotics literature, trajectory tracking using inverse dynamics and model reference control

has been used for some time now, and the methods developed therein can be seen as particular

subclasses of the formulation discussed in the present paper. Trajectory tracking in the adaptive

control context (which is not the subject of this paper) has also been explored together with

specific parameterizations to guarantee linearity in the unknown parameters of a system.

Controllers that Cause a Robotic System to Track a Given Trajectory

The equations of motion for constrained mechanical systems are based on the principle of

Lagrange. The principle states that at each instant of time t, a constrained mechanical system

evolves in such a manner that the total work done by all the forces of constraint under any set of

virtual displacements is always zero. This principle, which in effect prescribes the nature of the

forces of constraint which act upon a mechanical system, has been found to yield, in practice,

adequate descriptions of the motion of large classes of mechanical systems, thereby making it an

extremely useful and effective principle.

This paper takes a generally different approach that is based on recent results from analytical

dynamics. Here the complete nonlinear problem is addressed with no assumptions on the type of

controller that is to be used.

One considers the robot dynamics model, given by the joint-space formulation, usually presented

in the canonical forms:

FJqgqqCqtqMT)(),(),( (1.1.1)

M is an n by n symmetric, positive-definite matrix and is called the generalized, or joint-space,

inertia matrix, C is an n by n matrix such that C q is the vector of Coriolis and centrifugal

terms - collectively known as velocity product terms- g is the vector of gravity terms and F is a

vector of forces exerted by the end-effectors. More terms can be added to this equation, as

required, to account for other dynamical effects (e.g., viscous friction).

The symbols q, q , q , and τ denote n-dimensional vectors of joint position, velocity,

acceleration and effort variables respectively, where n is the number of degrees of motion

freedom (DoF) of the robot mechanism.

Page 18: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

18

This equation shows the functional dependencies explicitly: M is a function of q, C is a function

of q and q , and so on. Once these dependencies are understood, they are usually omitted.

Consider an unconstrained nonlinear mechanical robot system described by the second order

differential equation of motion under the form:

),,(),( tqqQqtqM (1.1.2)

0)0( qq 0)0( qq (1.1.3)

where, q0 and 0q are the position and velocity vectors at initial time of the robot with n DoF; the

dots indicate differentiation with respect to time. Equations (1.1.1) and (1.1.2) can be obtained

using Lagrangean model.

The n- nonlinear vector ),,( tqqQ , on the right hand side of equation (1.1.2) is a „known‟ vector in

the sense that it is a known function of its arguments. By „unconstrained‟ one means that the

components of the initial velocity 0q of the robot system can be independently assigned.

By „unconstrained‟ one means here that the n coordinates, q are independent of one another, or

are to be treated as being independent of each other.

One requires that this mechanical system be controlled so that it tracks a trajectory that is

described by the following consistent set of m equations:

0),( tqiΦ hi .....1 (1.1.4)

and

0),,( tqqiΨ mhi ,....1 (1.1.5)

Suppose further that the unconstrained system is now subjected to the m constraints.

One assumes that the mechanical robot system‟s initial conditions are such as to satisfy these

relations at the initial time. The latter set of equations, which are non-integrable, is non-

holonomic.

In order to control the system so that it exactly tracks the required trajectory i.e. satisfies

equations (1.1.3) and (1.1.4) one must apply an appropriate control n-vector ),,( tqqQ so that the

equation of motion of the controlled system becomes

),,(),( tqqQqtqM + ),,( tqqQc (1.1.6)

0)0( qq 0)0( qq

where now, the components of the n-vectors q0 and 0q satisfy equations (1.1.4) and (1.1.5) at the

initial time, t = 0.

Throughout this paper, one shall, for brevity, drop the arguments of the various quantities, unless

needed for clarity.

The controlled system is described by the relation (1.1.6), where Qc is the control vector.

Page 19: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

19

One begins by expressing equation (1.1.6) in terms of the weighted accelerations of the system.

For any positive-definite n by n matrix P (q, t), one define the matrix:

12/1 )],(),([),( tqMtqPqtqG (1.1.7)

Pre-multiplying equation (1.1.6) by P1/2

(q, t), the ‟‟pondered‟‟ equation, which will indicate

using the superscript p, is obtained as:

p

c

ppqaq (1.1.8)

where

aGap 1 , and c

p

c qGq 1 (1.1.9)

One designates the acceleration of the uncontrolled system by:

),,(),(),,( 1tqqQtqMtqqa (1.1.10)

In equation (1.1.6), one identifies the expression:

),,(),(),,( 1tqqQtqMtqqq cc

(1.1.11)

witch can be viewed as the deviation of the acceleration of the controlled system from that of the

uncontrolled system.

From equation (1.1.8), one obtains the expression:

cqaq (1.1.12)

One differentiates equation (1.1.4) twice and equation (1.1.5) once with respect to time t, giving

the set of equations:

),,(),,( tqqbqtqqA (1.1.13)

where A is an m by n matrix of rank k and b is an m-vector. With equations (1.1.6), (1.1.8) and

(1.1.12) equation (1.1.13) can be further expressed as:

),,(),,( tqqbqtqqB c (1.1.14)

where B is an m by n matrix who is calculated by the expression :

12/1 )],(),()[,,(),,( tqMtqPtqqAqtqqB c

(1.1.15)

One can now express the accelerations n-vector q in terms of its orthogonal projections on the

range space of BT

and the null space of B, so that:

qBBIqBBq )( (1.1.16)

Page 20: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

20

In equation (1.1.16), the matrix B denotes the Moore–Penrose generalized inverse of the

matrix B . It should be noted that equation (1.1.16) is a general identity that is always valid since

it arises from the orthogonal partition of the identity matrix )( BBIBBI .

Using equation (1.1.14) in the first member on the right hand side of equation (1.1.16), and

equation (1.1.12) to replace q in the second member, one gets:

))(( cqaBBIbBq (1.1.17)

which, due to equation (1.1.11), yields:

)( BabBqBB c (1.1.18)

The general solution of the linear set of equations (1.1.18) is given by:

zBBBBIBabBBBqc )]()([)()( (1.1.19)

After any combination one obtains the second equality:

zBBIBabBqc )()( (1.1.20)

where the n-vector ),,( tqqz is any arbitrary n-vector. To obtain the second equality above, one

used the property that )()( BBBB in the two members on the right hand side along, with the

property so that BBBB .

The set of all possible controls ),,( tqqQc (or controllers) that causes the controlled system to

exactly track the required trajectory is explicitly given by

zBBIPBabBPqPtqqQ cc )()(),,( 2/12/12/1 (1.1.21)

The mechanical robotic system, described by the nonlinear Lagrange equation (1.1.1), is

explicitly controlled through the addition of a control, n-vector ),,( tqqQc , provided by equation

(1.1.21), in witch the n-vector z may be chosen still ensures that the description of the trajectory

given by equations (1.1.3) and (1.1.4) is exactly satisfied.

Explicit Equations of Motion for General Constrained System

The constrained mechanical system described by Eqs. (1.1.1)-(1.1.4) evolves in time in such a

manner that the total work done at any time, t, by the constraint force n-vector Qc under virtual

displacements at time t is given by (1.1.21).

The work done by the forces of constraints under virtual displacements at any instant of time t

can be expressed as:

Page 21: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

21

),,(),,( tqqCwtqqQwT

c

T (1.1.22)

where ),,( tqqC is an n-vector describing the nature of the non-ideal constraints, which could be

obtained by experimentation and/or observation. The virtual displacement vector, w(t), is any

non-zero n-vector that satisfies:

0),,( wqqA t (1.1.23)

Solving equation (1.1.23), the n-vector w can be written as (we suppress the arguments for

clarity):

)( AAIw (1.1.24)

where γ is any arbitrary n-vector, and A+ is the Moore–Penrose inverse of the matrix A.

Substituting equation (1.1.24) in equation (1.1.22), one obtains:

),,()(),,()( tqqCAAItqqQAAI c

(1.1.25)

Since each component of the vector γ can be independently chosen, equation (1.1.25) yields:

),,()(),,()( tqqCAAItqqQAAI c

(1.1.26)

Pre-multiplying equation (1.1.6) by )( AAI and using equation (1.1.26), one gets

)],,(),,()[(

),()(

tqqCtqqQAAI

qtqMAAI

(1.1.27)

Equations (1.1.27) and (1.1.13) can now be written together as:

)],,(),,([)(),()( tqqCtqqQ

b

AAIqtqM

A

AAI (1.1.28)

Defining

),()( tqMA

A

AIM (1.1.29)

One can solve equation (1.1.28) to get:

Page 22: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

22

ηMMItqqC

b

tqqQtqMq )(),,(),,([),(

(1.1.30)

where is an arbitrary n-vector.

Equation (1.1.30) is the general, explicit equation of motion for constrained robotic systems with

non-ideal constraints. Here, we do not have any restriction regarding the positive definiteness of

the symmetric mass matrix M. It is allowed to be singular.

One can obtain the following results:

First result The general equation of motion of a constrained robotic mechanical system described

by relations (1.1.1)–(1.1.3), whether or not the matrix M that arises in the description of the

unconstrained motion of the system is singular, is given by Eq. (1.1.30).

The second member in equation (1.1.30) disappears for then TTMMMM

1)( .

In general, because of the second member on the right-hand side of equation (1.1.30), the

acceleration of a system with a singular mass matrix is not necessarily unique. However, when

the (m x n) by n matrix M

has full rank so that IMM

and so the equation of motion becomes

unique and consequently, one obtains another important result:

Second result When the matrix M

has full rank, the equation of motion of the constrained system

is unique and is given by the equation:

),,(),,([),( tqqC

b

tqqQtqMq

(1.1.31)

),,(),,([),()( tqqC

b

tqqQtqM

A

AAIq (1.1.32)

Conclusion

This paper presents the motion in terms of second-order differential equations. This methodology

has been inspired by results in analytical dynamics model.

The explicit closed-form expression (1.1.17) provides the entire set of continuous tracking

controllers that can exactly track a given trajectory description, assuming that the system‟s initial

conditions satisfy the description of the trajectory. The explicit closed-form expressions for the

controllers can be computed in real time.

Closed-form expressions for all the continuous controllers required for trajectory tracking for

nonlinear systems do not make approximations. Furthermore, no approximations or linearization

Page 23: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

23

are made here with respect to the trajectory that is being tracked, which may be described in

terms of nonlinear algebraic equations or nonlinear differential equations.

1.2 Design of flexible drive systems for feed-forward control: a mechatronic approach

Dynamic characteristics of a controlled mechanism result from choice of control law and

mechanical design at the same time, due to physical limits such as actuator saturations and

mechanical resonance frequencies. This paper concerns mechatronic design of mechanisms for

the feed-forward control of high-speed/precision motion. The proposed approach includes the

design of drive system and the motion control. The goal is to design both mechanical

transmission and electrical motor in order to achieve the performance specifications with the

controller.

The inverse elasto-dynamic model of the system is used at the same time to generate the 4th

order

feed-forward model for control and to establish the design constraints due to physical limits. The

design problem is then represented in the form of the Pareto front which allows a selection and

an optimal choice of the motor as well as an optimal sizing of the transmission mechanism. An

example shows that this design approach leads to obtain the performance specified at the

beginning of the process. Moreover, the performances are robust to modelling and identification

errors.

Introduction

The motion control of mechanism occurs in a broad variety of applications: industrial robot,

Computer Numerical Control machine tool, wing aileron motion control for airplane, drive

system for satellite attitude control and appendage maneuvers, etc. Three types of motion control

can be distinguished: target tracking (which the trajectory is unknown a priori), point-to-point

Page 24: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

24

positioning control (e.g. pick-and-place operations handling systems) and trajectory following

(e.g. laser cutting process).The command input for point-to-point positioning and trajectory

following is most often calculated according to a motion time evolution law. The intended

performance can come down to two objectives: agility (velocity and acceleration) and precision.

These ones are hard to reconcile due to elastic deformation and mechanical resonance. They

constitute an important problematic in the field of servomechanism control (Ellis and Lorenz,

2000), robots (Good et al., 1984), CNC machine tool, etc.

The contribution of this paper is of methodological order with an original approach and an

original formalism for drive system design. This paper is focused on position trajectory

following (linear and angular), according to a motion time evolution law, of a load with an

electric motor and a mechanical transmission. We propose an approach of mechatronic design

covering at the same time the drive system design and its control to obtain an optimum global

solution (plant and control).

A design global framework in mechatronic field is given for instance in the paper

(Cetinkunt,1991), but it does not exist any link between system design and its control. The

framework chosen here does not aim at the maximizing performance, but rather the answer to the

most right at a specific requirement by minimizing for instance the costs for industrial equipment

or the total mass of a space application. The performance requirements depend on the application

area; for instance the high-speed motion for robot manipulators, the precision for CNC machine

tools and the reliability in aeronautics.

So, to achieve the given performances (agility and precision), the mechatronic design depends on

a dynamic model of drive system by taking into account elastic deformations of the mechanism

and physical limits of such dynamic system. This model is used to generate a feed-forward

control.

The trajectory generation leans on a motion time evolution law which leads us to define the

motion specifications (agility) to be follow by the load. The model with the physical limits is

also used to define the design constraints of the drive system according to an appropriate

formalism. The selection and sizing method of the electromechanical components is succinctly

described. An example allows showing results obtained in accordance with the proposed

approach and, notably, to prove that the performances obtained with feed-forward control are

conform to the initial specifications and robust to modelling and identification errors. At first,

some details and reference points are provided on the proposed approach.

Preliminaries

The natural approach to design of a controlled system consists in separating the design of

physical system of the design of its control. This approach led us to have an over-sizing or

performances lower than specifications. The mechanism is indeed sized to transmit a mechanical

power (force, velocity). On top of that, to avoid any interaction with the control (resonance

excitation, spill-over phenomenon (Balas, 1978)) the mechanism can be over-sizing to place the

resonant frequencies high enough (of how much?), what leads to have a high mechanism inertia

and therefore an oversized motor.

In the case of the mechanism is not oversized (weaker inertia, lower resonance frequencies), a

control law damping the vibratory modes will be able to be used. A state space feedback control

Page 25: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

25

of flexible robot (Ryu et al., 2004) will prove to be often delicate to implement, what restricts its

use outside laboratories and space applications.

An attractive approach in set point tracking is the use of a feed-forward term established from

dynamic inverse model as it was proposed for the elastic servomechanisms (Piazzi and Visioli,

2000; Lambrechts et al., 2005), the elastic joint robots (Thümmel et al., 2005) and the flexible

link robots (Feliu et al., 2006). In addition, the authors of this last article offer a modification of

the dynamics of the arm of robot to be able to generate a simpler feedforward control.

Mechanical modification is obtained by addition of punctual masses along the arm in order to

achieve a relation specified between the frequencies of resonance (Feliu et al., 2006). These

works show interest to shape the system dynamics to achieve better results with control.

In this article, following the example of works (Feliu et al., 2006), an 4th

order feed-forward will

be established from inverse dynamic model including the rigid body mode and the first elastic

mode of the mechanism. We propose, as it was presented in introduction, to use this analytical

model for design, in order to include the predominant physical phenomena which will restrict the

performances of the control, as motors saturations and mechanical resonances. In brief, the

purpose is to adapt at best the drive system associated at its control to achieve the specifications.

The 4th

order elastic dynamic model is therefore defined at the same time for design and control.

Fig.1.2.1. Mechanism example

Dynamic model and physical limits

Let consider a load with inertia 2m and viscous friction coefficient

2c for the motion control

(linear or angular motion) such as the load position 2q must follow a set point cq t . The drive

system of the load (see Fig.1.2.1) is constituted by an electrical motor and an elastic

transmission.

Let us note for the motor: the rotor position1̂q , the motor force or torque f̂ , the inertial

parameters 1m̂ and the viscous friction

1̂c . The transmission mechanism is considered such as the

kinematic relation in the absence of elastic deformation is linear, i.e. 2 1̂q r q , where r is the

transmission ratio. The force transmitted by the mechanism, noted2f , causes an elastic

deformation of which the stiffness coefficient is noted k .

The dynamic model is described as follow:

q2

load

transmission

rotor of the

electrical motor

1q̂

Page 26: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

26

1 1 1 1 2

2 1 2

2 2 2 2 2

ˆ ˆ ˆ ˆ ˆ

ˆ( )

f m q c q r f

k q r q f

f m q c q

(1.2.1)

The instantaneous physical limits are imposed by the motor (its maximum velocity ˆmv and its

maximum force or torque mf̂ ) and the mechanism (its maximum acceptable deformation e ) such

as:

m

1

2 1

ˆ ˆf ( )

ˆ ˆ

ˆ

m

e

f t

v q t

q rq

( )

( )

( )

a

b

c (1.2.2)

Other limits, as the thermal limits of the motor or the technological limits peculiar to each

mechanism can be also considered.

In order to use a homogeneous model at the output 2q , the variables and parameters of the motor

are brought back to the load what defines new variables and parameters:

2 2

1 1 1 1 1 1ˆˆ ˆ ˆ, / , / , /q r q f f r m m r c c r (1.2.3)

The dynamic model can be rewrite as follow:

1 1 1 1 1 2( )f m q c q k q q (1.2.4)

2 2 2 2 2 10 ( )m q c q k q q (1.2.5)

Control scheme

The objective of follow-up trajectory cq t is obtained with a feedforward based on the inverse

dynamic model (IDM) of the elastic mechanism with the motor and a PD control loop to correct

the modelling and identification errors (cf. Fig. 1.2.2).

Fig. 1.2.2. Control scheme: FeedForward + FeedBack

The simplified diagram represented Fig. 1.2.2 leads to the motor force-setting FF FBf u u with:

mechanism

& load PD

IDM

qc

q2

uFF

uFB r

r

1q̂

Page 27: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

27

1 1( ) ( )FB p c d cu g q q g q q (1.2.6)

( )

1

( )n

i

FF i c

i

u t p q

(1.2.7)

The feedforward setting is therefore generated from inverse dynamic model of order n written

according to the output2q :

( )

2

1

( )n

i

i

i

f t p q

(1.2.8)

This feedforward requires therefore, on one hand, that the position set point cq t is n-1

continuous derivatives and, on the other hand, that dynamic parameters ip are identified on-the-

top. We restrict anticipation to 4th

order in accordance with the model introduced in section 3.

By tacking equation (1.2.5), it is possible to give 1q in terms of

2q and its successive derivatives:

2 2

1 2 2 2

m cq q q q

k k (1.2.9)

By substituting from equation (1.2.9) q1 and its derivatives in expression (1.2.4), we obtain the

dynamic inverse model (1.2.8) to 4th

order with the following parameters:

1 2

1 1 2 2 1 2

1 2 2 1 1 2

3 4

,

,

c cp c c p m m

k

m c m c m mp p

k k

(1.2.10)

The performance-robustness in presence of modelling and identification errors will be introduced

on an example in section 8.

Setting generation

The reference variable for point-to-point positioning control and trajectory following is most

often calculated by a planning algorithm for which the trajectory is free in the first case, imposed

in the second.

In both cases, the setting generation rests on a motion time evolution law which allows

generating a reference variable of amplitude displacement of q for the durationT :

0 . ( )cq q q Q t with t[0, T] (1.2.11)

where the set point cq t evolves from 0q to 0t and 0q q to t T ; Q is an interpolation

function advances between 0 and 1 and at least a function of class 3.

Page 28: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

28

Some examples of motion time evolution laws can be given. The first one is very practical, for

all the order of feedforward, and is defined by the following velocity law (B):

)T/t(sinT/2Q 2 .

A motion time evolution law in position defined by 2 polynomials of degree 4 with constant

velocity phase and boundary conditions (initial and final positions) of velocity and acceleration

set to zero (Khalil and Liegeois, 1984). If the initial acceleration constraint is relaxed, it is

possible to define an asymmetrical motion time evolution law by the polynomial of degree 4 (D): 432 386Q with t

T .

Aim of comparing these ones later, one can note the maximum magnitudes of velocity,

acceleration and jerk derivative:

(4)

max

max

max

c

c

c

v q

a q

b q

(1.2.12)

As well as the homogeneous variable according to a frequency (first harmonic approximation of

the motion time evolution law) such as:

b a (1.2.13)

The variables v , a and used to characterize the motion time evolution law will be justified by

equation (1.2.14) in following paragraph. These ones are used as design specifications of drive

system and constitute dynamic solicitation indicators for a given movement. For instance, let us

compare a bang-bang acceleration law – which includes a constant velocity phase – with the

motion time evolution laws introduced before. A displacement of 1 m must be achieved in 1 s,

initial and final velocities being set to zero.

The Tab. 1.1 provides indications on the motor solicitation ,a v and transmission

solicitation .

Tab.1. Motion time evolution laws characteristics

q

= 1 m

T = 1 s

(A)

bang

-bang

(B)

velocity in

sinus²

(C) two

polynomials

of degree 4

(D) one

Polynomial

of degree 4

v

(m/s)

2 2 2 1.7

a

(m/s²)

4 6.2 6 12

(Hz) 6.2 5.6 2.4

Of course, the bang-bang acceleration law is not usable for the 4th

order feedforward (i.e.

), but this law is the one which solicits the least the drive system in acceleration. We can

Page 29: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

29

point out for instance that law (C) is less demanding than the law (B). Law (D) is less demanding

for the mechanism than the law (C) but not for the motor.

Formulation for the design

The dynamic inverse model is used for the drive system design by neglecting the terms 1 2 /c c k

and 1 2 2 1( ) /m c m c k very small in front of the other terms, the stiffness k being necessarily high

and the friction coefficient 2c is assumed small. In addition, the inertial parameters

1m̂ , the motor

viscous friction 1̂c and the transmission ratio r are expressed in the model from relations (3) to be

able to size the motor and the transmission mechanism:

(4)1 2 1 1

2 2 2 2 2

ˆ ˆ ˆˆ m m m cf q r m q r c q

k r r

(1.2.14)

The characteristics of the motion time evolution law v , a and , equations (1.2.12) and

(1.2.13), are used with physical limits (1.2.2) to express the design constraints. By taking back

the motor force or torque limit (1.2.2a) with equation (1.2.14), the first design constraint

(1.2.15a) can be expressed. The second constraint (1.2.15b) is obtained with ˆmv , the velocity

limit (1.2.2b), and the kinematics relation of the mechanism 2 1̂q r q without considering its

elastic deformation. The third constraint (1.2.15c) concerns the maximum acceptable

deformation e of the mechanism (1.2.2c) expressed in terms of characteristics of the motion

time evolution law thanks to equation (1.2.7).

1 1 2

2 2

1 2

1

2

2 2

ˆ ˆˆ 1ˆ

ˆ

ˆ /

m

m

e

m m mf r m a

r k m r m

cr c v

r

v v r

m ca v

k k

( )

( )

( )

a

b

c

(1.2.15)

These three design constraints establish a relation between:

- Performance specifications: 2m , v , a and .

- 4 linked parameters (not independent) for the motor: ˆmv , mf̂ ,

1m̂ and 1̂c .

- 3 linked parameters for the transmission: r , k , e .

To be able to select and to size the components, it is necessary to regroup these linked

parameters, in inequations (1.2.15), in characteristic variables of components. The optimum

transmission ratio optr (optimum in term of acceleration) is introduced to show an adimensional

form with transmission factor r (Pasch and Seering, 1984):

Page 30: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

30

*

1 2ˆ/ /opt optr r r with r m m (1.2.16)

The choice of a factor * 1r maximizes acceleration for a motor given. By taking back the

formulation principle established for the choice of a motor and a transmission ratio (Dequidt et

al., 1999), a collection of parameters is achieved thanks to a metric defined by the motor and

load kinetic energies, respectively 1Ec and

2Ec :

2

1 1 1 1 1

2

2 2 2 2 2

1ˆ ˆ

2

1

2

Ec x with x m q

Ec x with x m q

(1.2.17)

The motor power output, with effective force or torque 1 1ˆ ˆ ˆ ˆ( ) ( )ef t f t c q , is rewritten with

1x to

show new variable F t :

1 1

1 1

1

ˆ ˆ ˆ( )ˆ ˆ( ) ( ) ( )ˆ

e

f t c qf t q F t x F t

m

(1.2.18)

Two homogeneous variables 1x and

2x (SI unit: J1/2

) are introduce and their temporal derivative

1x and 2x are homogeneous to F (t) (SI unit: J

1/2 s

-1), what allows us to rewrite the constraints

(1.2.15) under an adimensionnal form:

1* *2 1 * *

*

* *

11 1 ( )

( )

(1 *) ( )

a

c

ciné

e

Cr r r f a

C r

r r b

f c

(1.2.19)

With the motor constant aC =max(|F(t)|), the load constant

cC =max(|2x |), the minimum

transmission factor *

cinér , the external force factor (here the friction force applied to the load) *f ,

the transmission stiffness factor with the anti-resonance first pulse 2/z k m of mechanism

and the minimum stiffness factor e , such as:

1

1

ˆ ˆ ˆ

ˆ

m m

a

f c vC

m

2cC m a

2*

1ˆ ˆ

ciné

m

m vr

m v * 2

2

c vf

m a ,

2

2

z

,

2

1e

e

a

,

(1.2.20)

Page 31: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

31

Fig. 1.2.3. Design constraints (1.2.19a)

If we take for instance the constraint (1.2.19a), optimum is defined by a Pareto front (see Fig.

1.2.3) in a three-dimensional space; every axle takes back one or two components characteristic

variables: the ones of the mechanism, and *r , and the ones of the motor, aC and

*

cinér (constraint 1.2.19b). The graphs representing constraints (1.2.19) are plotted using a null

value of the friction coefficient 2c ( * 0f ).

Selection and component sizing

Constraints (1.2.19) can therefore be used to perform an optimum selection and sizing method of

the electromechanical components, either from an automatic minimization method (criterion:

mass, cost), or from an interactive method. These one can be made such as the designer is able to

choose the most suitable solution closed to the Pareto front (optimum solutions set). We offer in

this paper to describe a two-step design method (the selection of the motor and then the sizing of

the mechanism) from graphs defined with constraints (1.2.19).

From graph of Fig.1.2.4 (torque lower limit a cC C r for different minimum values of ), it is

possible to choose a motor by knowing the consequences of size on stiffness as well as to the

feasible range sup,cinér r of the transmission factor r (the line in bold for 4 , see Fig. 1.4).

Once the motor is chosen (motor load factor 2.5a cC C ), it is possible to size the transmission

by referring to graph of Fig. 1.2.5. The minimum stiffness factor is pointed out according to

the transmission factor r . On the other hand, the minimum stiffness factor corresponding to the

maximum distortion e is represented for solicitations on the motor side ˆe and on the load

side e .A feasible domain (pointed out D on the Fig. 1.2.5.) is defined for the choice and the

sizing of the mechanism components.

undersized

results optimum

oversized

c

a

C

C

r*

f*=0

Page 32: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

32

For an example of drive system design, the reader could see some works found in the literature

(Valdès et al.).

Let us take the example of a load of 120 kg moving according to the velocity law in sinus² (B),

cf. design specifications in the Tab.1.2.

Furthermore the dynamic behavior of the drive system which satisfies the design constraints

(1.2.19) is simulated with the proposed feed-forward control law.

Tab.1.2. Design specifications

m2

(kg)

Δq

(m)

T

(s)

v

(m/s)

a

(m/s²)

(Hz)

120 0.05 0.1 1 31.42 62.83

The dynamic behavior of the drive system with the 4th

order feed-forward control is compared

with a control without feed-forward (PD controller), cf. Fig. 1.2.6, 1.2.7 and 1.2.8. The Fig. 1.2.6

shows the setting and responses functions in position with and without feed-forward.

With feed-forward, the tracking error (cf. Fig.1.2.8) reaches a maximum of 2.6*10-4

m (0.5%

de q ), while it is 6.5 times bigger (1.67*10-3

m) for a control law without feed-forward.

0.5

1

4

f*=0

Fig. 1.2.4. Choice of the motor: constraints (19a, b)

r*ciné r

*sup r

*

c

a

C

C

5.2C

C

c

a

f*=0

e

e

D

Fig.1.2.5. Transmission sizing (1.2.19a, b, c)

r*ciné r

*

Page 33: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

33

When an identification error of 50 % is made on the load (2 21.5m m ), the tracking error is about

twice bigger (5.5*10-4

m); this error is equal to 3.5*10-4

m when the stiffness is divided by 10

( /10k k ).

In brief, the performances are robust to identification errors. Moreover, the design constraints

allowed (1.2.19) to size correctly the mechanism.

Let us notice that an overestimate of maximum motor force is imposed by the control law: 10254

N instead of 6340 N (cf. Fig.1.2.7). This is due to the "pessimist" approximation given by the

extreme (1.2.12) of the dynamic solicitation.

Fig. 6. position set point qc and load response q2 without and with

feedforward

t (s)

O1 q_c - loi de mouv ement sinus²O1 q_c - loi de mouv ement sinus²Displacement - m2 - av ec anticipationDisplacement - m2 - av ec anticipationDisplacement - m2 - sans anticipationDisplacement - m2 - sans anticipation

0

0.01

0.02

0.03

0.04

0.05

0.06

0

0.01

0.02

0.03

0.04

0.05

0.06

--mmmm

0 0.05 0.1

s

0 0.05 0.1

s

qC, q2 (m) Set point qc (motion law: sinus²)

Load position q2 with feedforward

Load position q2 without feedforward

t(s)

6340 N

Signal Output - u anticipSignal Output - u anticipSignal Output - u regulSignal Output - u regulSignal Output - u regul - sans anticipationSignal Output - u regul - sans anticipation

-10000

-5000

0

5000

10000-

-10000

-5000

0

5000

10000-

0 0.05 0.1

s

0 0.05 0.1

s

Fig. 1.2.7. Comparison between motor force command signals for feedforward + feedback

control

and for feedback control (without feedforward)

uFF, uFB (N)

t (s)

uFF uFB

uFB with uFF=0

f=uFF+uFB

Page 34: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

34

Conclusion

A mechatronic design method of flexible drive systems dedicated to feed-forward control has

been proposed. The originality of the approach is to enter at the same time the question of design

and the control to reach performance requirements. This approach uses the inverse elastic

dynamic model of the drive system and its physical limits (motor saturations, mechanical

resonance, etc.). It results at one and the same time a 4th

order feed-forward control law and the

design constrain of the drive system written under an adimensional form. The optimum design

problem is represented by a Pareto front (optimum solutions set) in a three-dimensional space.

Based on this formulation, a two-step design method has been established and applied to an

example in order to show the characteristics (performances and robustness) of a design solution.

One of the extensions of this approach of total optimization concerns nonlinear mechanisms and

arms of flexible robot with, in this case, a particular attention to non minimal-phase behaviour.

Acknowledgments

This work was supported in part by International Campus on Safety and Intermodality in

Transportation, the European Community (through the FEDER European Funds for Regional

Development), the Délégation Régionale à la Recherche et à la Technologie, the Ministère de

l‟Enseignement supérieur et de la Recherche, the région Nord Pas-de-Calais and the Centre

National de la Recherche Scientifique (CNRS): the authors gratefully acknowledge the support

of these institutio

Signal Output - écart position - sans anticipationSignal Output - écart position - sans anticipationSignal Output - écart position - av ec anticip. nominaleSignal Output - écart position - av ec anticip. nominaleSignal Output - écart position - av ec anticip. erreur chargeSignal Output - écart position - av ec anticip. erreur chargeSignal Output - écart position - av ec anticip. erreur raideurSignal Output - écart position - av ec anticip. erreur raideur

-0.002

-0.0015

-0.001

-0.0005

0

0.0005

0.001

0.0015

0.002-

-0.002

-0.0015

-0.001

-0.0005

0

0.0005

0.001

0.0015

0.002-

0 0.05 0.1

s

0 0.05 0.1

s

= qC-q2 (m)

1.67*10-3

3.5*10-4

5.5*10-4

2.6*10-4

max(||)

t (s)

Position tracking error without feedforward (FF)

Position tracking error with FF & without identification error

Position tracking error with FF & load identification error

Position tracking error with FF & stiffness identification error 10/~

5.1~22

kk

mm

Fig. 1.2.8. Comparison between position tracking errors without and with feedforward

Page 35: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

35

Chapter 2 Haptic interfaces

A haptic interface is a robotic system that creates communication between a human operator and

a virtual environment. It allows users to manipulate objects in a virtual environment with the

sense of touch and the kinesthetic perception. However, unlike the robot manipulators, the haptic

interfaces should have both high force rendering and maximum dynamic transparency (low

inertia, low friction, etc.). Therefore, the stability of haptic interface is a key issue. Any unstable

behavior arising during dynamic interaction may either damage this device or even injure the

human operator.

2.1 Haptic system modeling and stability analyze

Haptics is a enhancement to virtual environments allowing users to “touch” and feel the

simulated objects with which they interact. Haptics is the science of touch. The word derives

from the Greek haptikos meaning “being able to come into contact with”. The study of haptics

emerged from advances in virtual reality. Virtual reality is a form of human-computer interaction

(as opposed to keyboard, mouse and monitor) providing a virtual environment that one can

explore through direct interaction with our senses. To be able to interact with an environment,

there must be feedback. This type of feedback is called haptic feedback.

Haptic devices (or haptic interfaces) are mechanical devices that mediate communication

between the user and the computer. Haptic devices allow users to touch, feel and manipulate

three-dimensional objects in virtual environments and tele-operated systems. Haptic devices are

input-output devices, meaning that they track a user's physical manipulations (input) and provide

realistic touch sensations coordinated with on-screen events (output).

In human-computer interaction, haptic feedback means both tactile and force feedback. Tactile

feedback allows users to feel by her skin things such as the texture of surfaces, temperature and

vibration. Force feedback reproduces directional forces that can result from solid boundaries, the

weight of grasped virtual objects, mechanical compliance of an object and inertia.

Haptics has great potential in education-feeling is believing-especially in science and technical

areas, where forces are very important, being able to interact with a computer simulation and

literally feel the results. Students can develop a feel and appreciation for a particular subject even

before they have the detailed mathematics background to handle the equations.

INTRODUCTION

Nowadays, there are two main approaches to study the stability of a haptic device. In the first

approach, some authors have modeled the haptic interface like a sampled data system. In this

approach, the virtual environment is as a part of a model which consider the virtual impedance as

a linear spring and damper. The stability region is represented by the stability boundaries in the

stiffness-damping plan of the virtual impedance.

Hereby, the influence of parameters such as physical damping and time delay on stability

boundaries is evaluated with the transfer function model. The human arm impedance can be

included in the model to study the human dynamic effects. In the second approach, the theory of

two ports network is applied to separate the virtual environment and the human operator into two

Page 36: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

36

independent parts. Then, the virtual coupling is used and the stability is ensured through the

passivity or the unconditional stability criteria.

There are a lot of works concerning the interaction of the human operator, the haptic device and

the virtual environment. However, there remain still many challenges such as expanding the

boundary of stable region, modeling of human operator, designing of a control law, etc. In

previous studies, haptic device was modeled as a mechanical system including a mass or an

inertia, and a physical damper.

The existence of internal vibration of haptic device has been studied in recent years, but these

researches have not clearly explained the influence of the vibration mode parameters. The goal

of this paper is to analyze the influence of the dynamic parameters on the stability boundaries

based on the first approach. This study shows how the factors such as physical damping

coefficient, time delay and human operator parameters affect the stability of a haptic system. The

influence of vibration mode on the stability is also studied in this paper. The results are carried

out on the physical parameters of a PHANTOM®.

Haptic interfaces are relatively sophisticated devices. As a user manipulates the end effector, grip

or handle on a haptic device, encoder output is transmitted to an interface controller at very high

rates. Here the information is processed to determine the position of the end effector. The

position is then sent to the host computer running a supporting software application. If the

supporting software determines that a reaction force is required, the host computer sends

feedback forces to the device.

Actuators (motors within the device) apply these forces based on mathematical models that

simulate the desired sensations. One haves been developing haptic interface devices to permit

touch interactions between human users and remote virtual and physical environments. Examples

of haptic devices include consumer peripheral devices equipped with special motors and sensors

(e.g., force feedback joysticks and steering wheels) and more sophisticated devices designed for

industrial or scientific applications (e.g., PHANTOM haptic device, FALCON haptic device).

Phantom haptic device

The Phantom is a convenient desktop device which provides a force-reflecting interface between

a human user and a computer. Users connect to the mechanism by simply inserting their index

finger into a thimble. The Phantom tracks the motion of the user‟s finger tip and can actively

exert an external force on the finger, creating compelling illusions of interaction with solid

physical objects. A stylus can be substituted for the thimble and users can feel the tip of the

stylus touch virtual surfaces.

The Phantom Premium 6DOF is a desk mounted force feedback system that provides six degree-

of-freedom force and torque feedback. The system consists of the device itself, accompanying

power electronics, a remote/safety switch, and a PCI interface card. The Phantom Premium 3.0 is

a serial robot with 6 DOF. The device is leveraged from haptic technology developed and

implemented in standard Phantom force feedback devices [Fig. 2.1].

The Phantom Premium 3.0 is a large haptic device that accommodates free arm and hand

motions about the shoulder as the center of rotation. The Phantom Premium 3.0 base uses cable

capstan drive trains. Its motors are equipped with co-located position sensors. The base motor

(primarily responsible for left-right motions or x motions) is grounded. A cable transmission

converts motor rotations to the rotation of a large disk which carries the rest of the mechanism.

Page 37: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

37

The second and third motors (primarily responsible for up-down and in-out motions, or y and z

motions) ride on a ring mounted on the large disk, using a cable-capstan drive. The second and

third motors control the position of two linkages in a four-bar parallel linkage design. The

sensor-motor packages measure the endpoint position and provide force feedback in three

translational degrees of freedom.

The end effector takes the shape of a handle and measures less than an inch in diameter. It is

roughly the size and shape of a large permanent felt tip marker. It sports one switch which can be

programmed to produce visual and/or haptic effects.

Fig. 2.1.1 Phantom Premium 6DOF haptic device

The 6DOF device is driven by a 6-axis power amplifier box and interfaces to a Pentium based

Windows NT computer via a PCI controller card. Low level communications between the PC

and the PCI card are handled by the Phantom Device drivers (PDD). The PDD maintains a 1 kHz

servo update rate to ensure stable closed loop control of the device. The device kinematics and

other robotic calculations are handled within the PDD. High level force and position calculations

are provided by the the Software Developer's Kit (SDK). The SDK provides a high level C++

programming interface for generating haptic effects.

Haptic effects handled by the SDK can be based on geometry (such as point haptic exploration),

or on force-time profiles (such as sinusoidal vibrations and jolts). The user can define their

owned custom force fields. Currently the SDK only supports 3DOF forces based on point haptic

exploration. Using new extensions allow a 6DOF application to read the Denavit-Hartenberg

homogeneous transform, describing the global position and orientation of the end effector,

calculate 6DOF forces and torques in the application, and command these 6DOF global forces

and torques to the device.

Falcon haptic device

The Novint Falcon is a pointing device with three arms coming out of the main body of the

device. Each arm is capable of pulling and pushing, and is updated every 1/1000 of second. In

addition to the arms there is an interchangeable handgrip, with three programmable buttons on it.

Page 38: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

38

Fig. 2.1.2 Novint Falcon haptic device

Falcon haptic device is a Delta parallel robot. Parallel mechanisms are a natural fit for haptics,

because they offer good stiffness and robustness at a relatively low cost. Stiffness in particular, is

important in haptics, because motors can‟t respond instantaneously.

A grip's position in real space corresponds to the location of a cursor in the three-dimensional

virtual world. And when that cursor touches a virtual object, the Falcon device sends an

analogous force back to the grip. Likewise, forces that the user exerts on the grip are translated

into cursor movements.

The Falcon haptic device generates its real world forces with three small joystick-sized motors

connected to the device's three arms, by a cable linkage. The system can produce a force in any

direction, within its 4 x 4 x 4 inch working envelop.

Novint's control software for the Falcon adjusts the motor current - in accordance with what's

happening in the virtual world - with textbook physics equations, determining the magnitude and

vector off all the forces and deflections. The system updates current positions - real and virtual -

at 1 kilohertz. Falcon system achieves a sub-mm resolution.

The Falcon may be getting its start as a game controller, but the Falcon isn't just one device, it's a

haptic platform that could be applied more broadly, as a 3D input/output device for a variety of

computing systems, including design and analysis software used by engineers.

HAPTIC SYSTEM MODELING

A haptic device is composed of two principal parts: the electromechanical part – containing

actuator and mechanical transmission – and the controller part – which is a link between the

physical device and the Virtual Environment (VE). Moreover, there are two challenges in the

design and control of haptic device. The first challenge is to guarantee the stability when a user

manipulates an object put in contact with a virtual wall (constrained motion, cf. Fig. 2.1.3).

Taking a high value for the virtual stiffness to represent a stiff wall can lead to an instable

behavior. The second challenge is the transparency of haptic device when a user manipulates an

object in a free area (i.e. when there is no contact with the virtual wall, cf. Fig.2.1.3). In theory, it

is advisable that the human operator don‟t feel the inertia and the viscous friction (damping) of

Page 39: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

39

the device in order to have no distortion of the force rendering during virtual object

manipulation.

In order to design and to control a haptic system, first of all, its basic model should be

established. A 1-DOF (degree-of-freedom) mechanical interface is usually modeled as a linear

rigid mechanism including a mass or an inertia (m), a physical damper (b) i.e. a viscous friction

(cf. Fig. 2.1.3).

The arm position X2 corresponds to the actuator position X1 without elastic deformation of

mechanical transmission: X2 = X1. Impedance-based virtual environment for constrained motion

is a coupling system including linear spring with virtual stiffness (K) and a virtual damping

coefficient (B). The resulting force F1 is applied by the motor on the mechanical transmission

which is opposed by the force F2 applied by the human operator.

Fig. 2.1.3 Free motion in VE: transparency challenge

Page 40: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Aurel FRATU

Habilitation Thesis

40

It is assumed that some nonlinear phenomena (like sensor resolution, actuator saturations, etc.)

are negligible. In particular, dry friction and sensor quantization can be ignored. The sampled-

data and the time delay due to the computer part (controller, virtual environment computation,

and communication) are included in the haptic interface model.

With the sampling process and the zero-order holder (ZOH), the overall model is a hybrid one

with both continuous and discrete forms. In the following, this model is used in a purely discrete

form, cf. Fig.2.1.4.

Fig. 2.1.4 Dynamic model of haptic interface and its discrete form

METHODOLOGY

In this part, two methods are presented to find the stable region of haptic device which are the

Routh- Hurwitz criterion and the gain margin. First, the characteristic equation of haptic system

is derived from the hybrid dynamic model (cf. Fig.2.1.4). The transfer function of haptic

interface is defined as:

Page 41: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

41

(2.1.1)

The rigid model of the mechanical interface G(s), the virtual interface GE(s), the ZOH H(s) and

the time delay D(s) are formulated as:

(2.1.2)

where Z[.] is the Z-transform operation of the function into bracket, and * is signed by discrete

signal. The characteristic equation of this system is:

(2.1.3)

Effect of human operator model

In this part, we study the influence of human operator to the stability of haptic system. The

human operator is usually modeled as a linear second order system including a mass MH

coefficient BH and a stiffness coefficient KH impedance is characterized by:

ZH (s)=MH.s +BH.s + KH (2.1.4)

This human model is coupled with the device model seen in Fig.2.1.5. In order to study the

influence of human operator on the boundary of stable region, we used parameters in Table II

which correspond to several models.

The Fig. 2.1.6 shows the stability boundaries ofthe PHANTOM® including human operator

models of some authors. As pointed out, these results show that human operator will be able to

stabilize the system and the smallest stable region is obtained without human model. Therefore,

in the worst case, the human operator part should be ignored in the model for stability analysis.

Page 42: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

42

Fig.2.1.5 Dynamic model of haptic system with human operator

Fig.2.1.6. Stability boundaries for human operator models

EFFECT OF VIBRATION MODE

In previous studies, a rigid model of haptic used to analyze the stability. In fact, there exists an

elastic deformation of mechanical transmission components between the motor shaft (position

X1) and the device arm (position X2). Consequently the resulting vibration mode can modify the

stability boundaries. A more complete model of haptic device with an internal vibration mode is

shown in Fig.2.1.7. The relationship between outputs positions and input forces for this haptic

device which has been defined by transfer function is presented here.

Similar to the Fig.2.1.4, the transfer function of sampled output position, X*1, and input force F

*2

are given by (2.1.1) with the transfer function G(s) defined by (2.1.4). Note that, in this part, we

only study the haptic system with vibration mode and without time delay.

Page 43: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

43

(2.1.5)

The critical virtual stiffness is determined with D(z)=1 (d=0). The gain margin method is used to

compute the stability boundaries for several different values of anti-resonant frequency wa.

Obviously the other physical parameters are considered to be constant. The results are based on

the physical parameters of the first vibration mode of the PHANTOM® (cf. Table III).

Matlab® is used to compute the gain margin. This method is very useful to reconstruct the

stability boundaries of haptic devices even if its transfer function is very complex. In contrast,

the first method becomes unusable while the order of characteristic equation is high.

Fig.2.1.7 Haptic device model with internal vibration mode

The Fig. 2.1.8 shows that the stability boundary is reduced when the anti-resonant frequency

decreases in the range of wa≤369 rad/s. From the Bode diagrams of the discrete transfer

functions Z [G(s).H(s)] with this range of anti-resonant frequency, as seen in Fig.2.1.9, it is

obvious that the higher gain margin of the system is obtained for the higher value of anti-

resonant frequency.

Page 44: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

44

Fig.2.1.8. Effect of anti-resonant frequency for wa≤369 rad/s

Fig.2.1.9 Bode diagram of Z [G(s).H(s)] for wa≤ 369 rad/s

This paper presents an overview on the effects of dynamic factors on haptic interface stability.

The impedance model used to compute the force feedback includes a virtual stiffness and a

virtual damping.

This paper has provided engineers some basic guidelines in choosing design specifications of

haptic system. In future works the effect of others physical parameters (e.g; time delay, human

model) will be studied with a model included vibration modes.

ACKNOWLEDGMENT

Page 45: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

45

This work was supported in part by the International Campus on Safety and Intermodality in

Transportation, the European Community (through the FEDER European Funds for Regional

Development), the Délégation Régionale à la Recherche et à la Technologie, the Ministère de

l‟Enseignement supérieur et de la Recherche, the région Nord Pas-de-Calais and the Centre

National de la Recherche Scientifique (CNRS). The authors gratefully acknowledge the support

of these institutions.

Page 46: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

46

2.2 Hybrid force/position control of the scissors-robot

In this paper a method for displaying feeling of cutting, using a virtual scissors type, is described.

A new computer haptics algorithm to be used in general interactive manipulations of deformable

objects is presented. The algorithm was developed to simultaneously display translational and

rotational (cutting) displacement for a realistic cutting simulation. We showed that interactive

animations can be used to simulate the functioning of force/torque - feedback scissors devices.

This paper discusses both the experience with the design process in the virtual environment and

the novel features of the possible portable torque-feedback scissors devices. We will present an

approach for designing a more natural user interface, without resorting to various special haptic

input/output devices.

Haptics are emerging as effective interaction aids for improving the realism of virtual worlds. To

be able interact with virtual environment; the user should be able to touch a virtual object and

feel a response from it. This type of feedback is called haptic feedback. The confluence of

several emerging technologies made virtualized haptics, or computer haptics, possible.

Much like computer graphics, computer haptics enables the display of simulated objects to

humans in an interactive manner. However, computer haptics uses a display technology through

which objects can be physically palpated. This new sensory display modality presents

information by exerting controlled forces on the human hand through a haptic interface (rather

than, as in computer graphics, via light from a visual display device). These forces depend on the

physics of mechanical contact. The characteristics of interest in these forces depend on the

response of the sensors in the human hand. Although these haptic interface devices generate

reactive forces when their users grasp or touch some objects with a device indirectly, the contact

type device represents a tactile sense by touching objects directly with finger.

Unlike computer graphics, haptic interaction is bidirectional, with energy and information flows

both to and from the user. The combination of high-performance force-controllable haptic

interfaces, computational geometric modelling and collision techniques and an understanding of

the perceptual needs of the human haptic system allow us to assemble computer haptic systems

that can display objects of sophisticated complexity and behavior. Therefore, they must allow an

as natural and intuitive as possible manipulation at distant of virtual objects in order to enable a

mental projection of the user in this universe.

In human-computer interaction, haptic feedback means both tactile and force feedback. Tactile,

or touch feedback is the term applied to sensations felt by the skin. Tactile feedback allows users

to feel things such as the texture of surfaces, temperature or vibration.

Force feedback reproduces directional forces that can result from solid boundaries, the weight of

grasped virtual objects, mechanical compliance of an object and inertia. Haptic technology

covers a wide spectrum of human interaction with virtual environment. Haptic devices allow

users to touch, feel and manipulate three-dimensional objects in virtual environments and tele-

operated systems.

Page 47: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

47

A haptic interface device is a device that provides an operator a tactile sense, and the operator

can manipulate objects in the three dimensional virtual environment and can acquire their feel of

touch. Haptic is the sense of touching something.

A haptic device is connected to a virtual world for enabling touch of three-dimensional modeled

objects. If there is a collision in the virtual world between a tool and the object, the device

provides force/couple feedback to the user by combining the advantages of computer (virtual)

models and physical models.

Interactive simulations, haptic feedback computation often comes from contact forces.

Subsequently, the fidelity of haptic rendering depends significantly on contact space modeling.

Contact and friction laws between deformable models are often simplified in up to date methods.

They do not allow a "realistic” rendering of the subtleties of contact space physical phenomena

such as slip and stick effects due to friction or mechanical coupling between contacts.

Some current researches are focused on the mastery of tactile interaction with distant physical

objects via virtual objects, due to haptic technology which may result in applications and

advanced manufacturing [1]. Actual technology enables us simultaneously to see and feel a

virtual object. Research in multi sensory human-computer interaction currently focuses on

making greater use of the senses of vision and touch.

Virtual environments seek to create computational scenarios that combine sensory data as sight

and touch to form an advanced, realistic, and intuitive user interface.

Using haptic interface, the user can directly manipulate computer-generated virtual objects with

his fingers and feel shape and texture [2]. In this system, user can operate virtual objects

naturally as in the real world.

The algorithm to generate a haptic response for scissors model

The algorithm would have to be developed and implemented to simulate the dynamic system

response. The haptic algorithm is composed of following steps:

building a proper virtual model of the scissors, namely the virtual scissors;

finding the closest points to a possible collision along the blade of scissors (collision points), both for self-collision or collision with external objects;

generating cutting/repulsion torque (or Cartesian velocities);

creating a small linear cutting force/torque ramp for different thicknesses of the cutting material;

computing the cutting torque (or joint velocities) commands, to be transferred to the real commands via the controller of the physical scissors device.

Virtual Prototype of a Pair of Scissors

Virtual reality is also known as virtual (or synthetic) environments and is widely used as a

training tool in engineering. Virtual environments are three-dimensional computer generated

environments that the user is able to experience interactively. The manner humans interact with

their physical environments, is artificially imitated in virtual environments. The modelling tool

provides a modelling environment that automatically generates a simulation script for the

dynamic response to user controlled conditions and variable values. The output data can be

viewed, exported or graphed with little human effort.

Page 48: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

48

Virtual environment provides a natural interface between humans and computers in virtual

reality. The interaction between virtual environment and physical environment (world) is

mediated by the haptic interface that allows to be manipulated in real time [3].

A haptic device consists of an interfacing system with humans through output of sensory

information and input of commands. An example of such input and output devices is the haptic

scissors.

The major goal of computer graphics and virtual environment simulation is to provide realistic

methods to allow us easy creation of digital equivalents for natural phenomena such as scissors

dynamic behavior and its interactions.

With up-to-date simulation tools one can deal with exact geometry, consider the dynamic

characteristics of a pair of scissors, include the man–machine interfaces, and visualize objects in

three-dimensional space in detail.

Having all this in mind and evaluating the possible solutions, the boundaries can be pushed

further away, especially when using advanced virtual reality tools. Simulation is employed to

predict cutting forces and dynamic response of the haptic system during different operations and

following adequate scenarios. Scenarios can serve as specifications for virtual prototypes. This

technique emphasizes narrative, rather than theory, as the primary source for designers.

Scenarios go to a high level of detail by creating an illustrated script, thus providing a more

detailed specification for prototype software.

Using simulators, we may build experimental environments, according to our own imagination.

What we have imagined during the night can be transposed into visual imagines the next day.

Complexity, specificity can be gradually increased to a level where virtual systems can lead to

real challenges of the physical world. The ability to simulate opens a wide range of options for

solving many problems creatively. We can investigate, design, visualize and finally see the

performances of a scissors in the virtual environment before it is used on real object. It is

possible that our solutions may fail or even blow up, but only in simulating on the virtual

prototype. It forces us to think thoroughly and understand how the system really works.

In this paper the authors propose the Delphi software environment, for the simulation of the

haptic systems, using visual programming. Delphi Simulators are the basic tools for creating the

virtual prototypes. They are suited for users who know the visual programming, but visualizing

still needs another demand. Delphi Simulators enable more open and flexible software tools for

programming the algorithms for haptic portable torque-feedback scissors.

The simulator created in the Delphi environment was used to test the performances of the haptic

scissors in their integrated environment. The simulator created with the DELPHI code, allows us

to create the virtual scissors prototype. This model deals with the characteristics of motion,

regarding the effects of forces or masses. The simulation methodology uses a number of possible

scenarios and involves the presence of the user to create some sequences of animation in the

virtual environment, as one can see in Fig. 1.

Page 49: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

49

Fig.2.2.1 Sequence of animation in virtual environment

For writing applications that produce three-dimensional computer graphics, we used Delphi

language and graphical objects existing in OpenGL standard library. The Delphi source code -

only one procedure - for this application, is shown below.

procedure DrawModel(M : TModel);

var I, J, K : Integer;

begin

For I :=1 to M.Groups do

begin

glMaterialfv(GL_FRONT_AND_BACK,GL_DIFFUSE,

@M.Material [M.Group[I].mIndex].Diffuse);

glMaterialfv(GL_FRONT_AND_BACK, GL_SPECULAR,

M.Material[M.Group[I].mIndex].Specular);

glMaterialfv(GL_FRONT_AND_BACK,GL_AMBIENT,

@M.Material[M.Group[I].mIndex].Ambient);

glMaterialfv(GL_FRONT_AND_BACK,GL_SHININESS,

@M.Material[M.Group[I].mIndex].Shininess);

glDisable(GL_TEXTURE_2D);

for J :=1 to M.Group[I].Faces do

begin

with M.Group[I].Face[J] do

begin

case Count of

3 : glBegin(GL_TRIANGLES);

4 : glBegin(GL_QUADS);

else

glBegin(GL_POLYGON);

end;

Page 50: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

50

for K :=0 to Count-1 do

begin

if M.Normals > 0 then

glNormal3fv( @M.Normal[nIndex[K]] );

glVertex3fv( @M.Vertex[vIndex[K]] );

end;

glEnd();

. . . . . . . . . . . . .

end;

Rendering of virtual Scissor capability

Using simulation, we can learn about some things in a very effective way and while modifying

“rules” we can observe the effects of our intervention. As it is in our nature that “seeing and

feeling is believing”, the visualization and haptic sensation are the dominant characteristics of a

haptic device [3].

During simulation a model is driven by input data and produces output data. Data can come from

initial values in the model, user entry or a data file. Output data can be stored in the model or in

output data files and used to drive configured charts. An integrated Data File editor makes it easy

to create, view, import data from or export data to other applications.

Concerning the cutting process, simultaneous rendering of cutting (rotational) and translational

movement is necessary for realistic simulation.

Virtual scissors can be handled using graphics modeling. We like the visualizing properties of

the virtual scissors, but we are alarmed about their complexity. We would also like to take

advantage of the knowledge about physical model, which is constrained mainly in favor of

virtual model.

Rendering of virtual Scissor Action

The geometric modeling cutting (rotational) movement is performed by computing the Boolean

operation between the tool object and the modeled work object [4]. The first step is recognition

of the direction of motion of the blades: opening or closing. This is determined by the sign of the

angular velocity of the scissor blades.

The subtle Boolean operation simulates the modeling process which deforms the shape of the

work object by adding the band on it. The shape of the object is modified interactively.

In the case of cutting, the object after modeling can be created by removing raster graphics

images (bitmaps) of the inside attribute of the work object and maintaining the inside attribute of

the work tool (scissors). An intuitive modeling method is possible by representing cutting

resistance as tactile sense information to the user. The algorithm to generate tactile sense

depends on the deformation modeling methods [5]. In the deformation modeling method, the

objects can be deformed and the user can touch them and feel twist force. In the non-modeling

method, the objects can not be deformed and the user can only touch them.

Page 51: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

51

The haptic rendering algorithm in the deformation modeling method generates the cutting force

proportional to cutting volume and a tactile force proportional to speed. It is difficult to calculate

the volume dislocated by the tool precisely in point-based geometry. Therefore we use the

number of raster to be removed during the cutting operation, as a dislocated volume.

The haptic rendering algorithm in the deformation non-modeling method generates only a tactile

force proportional to speed. The magnitude of the force is calculated by means of the empirical

expressions. The cutting force proportional to the number of raster restricts the withdrawal

movement of the scissors tool from the work object if it is always applied while the scissors tool

penetrates the work object. Therefore, only a speed element in the direction to penetrate the work

object is added to the reactive force as the resistance force. The direction of the reactive force is

given by the normal vectors on the surface of the work object.

For the calculation of the reactive force, only a speed in the penetration direction is calculated,

while the withdrawal speed, in the opposite direction, will be ignored.

The Scissors tool can be used to split paths, shapes and frames and to reshape them. We

proposed to realize the following tasks, with the virtual scissors tool, in the virtual environment:

cut straight across a shape;

cut a piece out of a shape;

cut a curved piece out of a shape;

cutting or modifying paths and frames.

Cutting straight across a shape with the scissors tool is one of the scissors‟ tasks. Imagine a

cardboard rectangle that we need to cut with the scissors tool in the virtual environment.

These were the steps followed:

first, we selected the scissors tool in relation to the thickness of the cardboard.

then, we moved the scissors till it collided with one side of the rectangle, right on the point where we needed to split paths,

then we clicked on the right mouse button and the scissors blades started to close.

We moved the left-hand portion of the rectangle away from the right-hand portion in order to

better observe the cut, as is presented on the Figure 1. The user moves the cursor in the virtual

environment freely using a device, and can operate spontaneously.

In the virtual environment we can also move the cardboard by using the A and D key on the

computer keyboard, for left/right direction or the W and S key, for top/bottom direction so as to

bring the cardboard into contact with the cutting edge of the scissors.

Rendering of virtual scissor orientation

We can rotate our item precisely by selecting it on scene that we want to rotate and then clicking

on the item, we drag away from the object if we want to rotate it clockwise, or drag towards the

object to rotate it counter-clockwise. To rotate the scissors‟ blades, for a positive value of the

angle of rotation the upper blade will rotate counter-clockwise, while for a negative value, it will

rotate clockwise.

Page 52: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

52

After selecting the item we want to rotate, with the mouse tool, we start to rotate the image frame

along with the image contained within. If we have selected our item, then our virtual tool will

rotate, but the frame containing it will not. To rotate the selected object, we must click the left

mouse button, hold down and subsequently drag the cursor. The rotation of the virtual object

reveals the silhouette of the object. Our attention is focused on the orientation of the scissors‟

blades and the direction of drawing the contour that must be cut.

Using visual controls would intuitively place the scissors in the desired position. This is useful

when trying to give the impression of something being closer or further away, when we want to

detach it from the opening cut into the object. Practical contour cutting was performed with

some small objects as an A4 cardboard. If we cut a frame that has an image in it, the image

contained in the frame won't split, but we will have a copy of the image in the frames we will get

after having applied the scissors tool. In the photograph below, we note how the dark lines on the

cardboard edge reveal the form that must be cut. Dark lines on the cardboard edge mark out the

pathway of contact points of the cutting edge of the scissors with the piece of cardboard, as can

be seen in Fig. 2.2.2.

Fig. 2.2.2. Pathway of contact points between the cutting edge and cardboard

Subsequently, the dark lines in the virtual environment must be reproduced by the haptic device

in the physical environment (world). The line that runs across an object, hinting at the cutting

form, is a cross-contour and is illustrated by the dark lines as is shown above.

In this example, our scenarios were simple enough so that the absence of detailed specifications

did not obstruct our later prototyping work. The model was parameterized with data and

equations and simulated over time. Simulation results can be viewed on a graphical monitor, and

finally exported to a physical device.

Design of a Portable Torque-Feedback Device

Today most force/torque-feedback interfaces come in the form of small robot manipulators:

these systems use electric motors to generate forces and joint sensors to measure spatial position

[6].

In the physical world the static state of deformation in a plastic object under load generally

depends on the forces that are applied to the boundary of the object [7]. When a rigid tool

contacts a deformable object, the position and orientation of the tool with respect to the object

uniquely defines the forces applied to the object. Therefore, the resultant force applied by the

tool can be expressed by a nonlinear function of the position and orientation of the tool.

Page 53: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

53

Haptics is enabled by small robot manipulators that apply the forces to the skin for touch

feedback. The actuator of the small robot provides mechanical motion in response to an electrical

stimulus.

In our application we used the electric devices that reproduce high fidelity tactile sensations.

There are a lot of devices adopting physical mechanisms in the tool assisted type. Our prototype

of a scissors-type haptic device consists of three components: an interface, a drive component,

and position sensors. The interface is designed to provide the features of a pair of scissors. The

drive component is composed of a DC-motor as actuator who generate reactive forces. We used

the Gill 360º Blade Rotary Sensor as position encoder.

System outline

The haptic scissors system implements the idea of virtual splitting using virtual scissors, while

the operator‟s hand is holding the physical scissors. The outline of the haptic system developed

in this research is shown in Fig.2.2.5.

The system consists of a physical scissors device that provides three dimensional position and

orientation outputs for a vectored tactile sense input, a control box for device control, a computer

and a modeling software system.

The feeling of cutting resistance and contacting the surface can be conveyed through a physical

device. The moving parts of the physical device - mobile blade - must be light-weighted to

provide outstanding operation feeling.

The operator grips the handle part of the physical scissors attached to the device, and so provides

input for the virtual system. The hand serves as a grasp, allowing operators to maintain the

model in its current position. Moving the physical scissors while pressing the blades does affect

the position and orientation of the associated virtual object.

The sense in which a user touches through the tool can be reproduced easily in comparison with

the sense in which a human touches with the finger; hence the scissors type device can reproduce

a better sense of reality by limiting its tactile sense range to the contact point of the moving blade

with the cutting object. This hand-held device would have one torque motor to provide torque

feedback: as the operator spins the shaft, the device can spin back as appropriate.

The physical cutting device developed in this research is equipped with a DC servo motor and a

rotary encoder. A current amplifier was used for DC servo motor to generate a torque of arbitrary

magnitude.

Our portable torque-feedback device uses the same basic electronics as standard version of the 2-

D joystick, but with additional multifaceted mechanics.

The position and posture inputs from the device to the computer are provided by the rotary

encoder. The rotation angle of the axis represents the input to the counter circuit and is applied to

the PC bus of the computer as a pulse signal, whereby a rotation angle is measured by counting

the number of pulses. The rotary encoder provides the signal for the operation of the positional

input block. The power supply to the DC servo motors is controlled by applying a digital signal

of variable frequency between 5 Hz and 5 kHz generated by the control circuit.

Page 54: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

54

In this research we used a DC servo motor with a nominal voltage of 18.0[V], and a breakdown

torque of 0.2 [Nm].

Sensor and actuator signals are applied to the supervisor system that also controls timing of the

system.

The signals are transferred - via a microcontroller – between the physical world and the virtual

world and by running the Delphi program on a PC, the virtual scenes are displayed on the video

monitor.

The common problem in the design of the haptic scissors is how to apply the fundamental laws

that govern the cutting process. The cutting dynamic model is known to be difficult to solve in

an efficient way, sometimes even too difficult to be solved at all. Another complexity is the fact

that a real-time rendering in immersive virtual environment, with haptic interaction, requires

high frequency processing (1000 Hz) and the visual aspects need frame rates in the order of 30

fps.

Because of the noise that arises when differentiating the position obtained from inductive

encoders, position data was passed through a first-order filter with a corner frequency of 10

rad/sec before differentiation. Even with filtering, however, the system still exhibited vibration

effects at low velocity.

This problem was overcome and the system was stabilized by creating a small linear force/torque

ramp-down band around zero velocity. Within this band, the cutting force/torque was multiplied

by a scale factor proportional to the velocity. The scale factor decreased as zero velocity was

approached from either positive or negative directions, and became identical to zero when the

velocity became zero.

This was possible due to an amount of force/torque output large enough for moving the scissors

blades in opposite directions, changing the sign of the velocity and consequently the direction of

the output force in a cyclical manner.

Hybrid force/position control of the scissors-robot

This section presents an analytical model based on the concepts of contact mechanics and

fracture mechanics to calculate forces applied to scissors during cutting of a plate of material.

The model considers the process of cutting as a sequence of deformation and fracture phases.

During deformation phases, forces applied to the scissors are calculated from a torque-angle

response model synthesized from measurement data multiplied by a ratio that depends on the

position of the cutting crack edge and the curve of the blades.

We will present in following the cutting process of a plate of material with a pair of scissors. We

will analyze two cases: only the upper blade of the scissors has angular motion during

interaction, in the first case and second case, when twice blades have angular motion.

Figure 3 shows cutting of a plate of material when only the upper blade of the scissors has

angular motion during interaction. The plate rests along the edge of the lower blade of the

scissors and it does not move during cutting.

Page 55: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

55

Fig.2.2.3 Explanation for a scissors with the upper blade angular motion

Only the pivot of the scissors can move on the surface of the plate along a straight line parallel to

an axis denoted by x. The lower blade of the scissors does not have any angular motion.

The coordinate xp defines the location of the pivot and θ defines the opening angle of the

scissors. The component of force applied to the upper blade in the direction normal to the blade

edge is needed for haptic applications. This component is defined as fn, and it is calculated

during two distinct regimes by a contact model.

Contact Model

The static state of deformation in an elastic object under load generally depends on the forces that

are applied to the boundary of the object [3].

When a rigid tool contacts a deformable object, the position and orientation of the tool with respect

to the object uniquely defines the forces applied to the object. Therefore, the resultant force applied

by the tool can be expressed by a nonlinear function of the position and orientation of the tool.

When we cut an object with a pair of scissors, the force that we feel between our fingers includes

two main components: friction forces of the contact of the blades, and the forces of cutting of the

object. In this approach we do not model the friction forces. However, friction forces can be

measured by opening or closing empty of the scissors, and be summed into the model.

Figure 4 shows the interaction between a pair of scissors and a thin plate with thickness h at the

time t when the plate is locally deformed.

A Cartesian frame is defined at the pivot of the scissors such that the x axis is along the symmetry

line of the scissors. We assume that the pivot of the scissors does not change orientation during

cutting.

We also assume that the pivot of the scissors does not move. The twice blades have angular

motion. The opening angle of the scissors is defined by θ, and the position of the edge of the crack

made by the scissors is defined by the point xc , with respect to the pivot point of the scissors. The

blades locally deform an area of the plate around the crack edge. The deformation can take

various forms including twisting, stretching, compression, or a combination there of.

Page 56: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

56

Fig.2.2.4. The states of scissors and a plate during the cutting process

At time t, the plate is locally deformed. During time period dt, a small area of the plate, h dxc, is

cut.During deformation, the upper edge of the rupture point is displaced from (xc, h/2) to (xc, h/2

− δ), where δ is a displacement length.

In response to deformation of the plate, the force fn is applied to the upper blade along the

normal to the blade‟s edge at point (xc, h/2 − δ).

The displacement length δ is determined on the base of the curve of the edge of the upper blade.

The curve of the blade edge has a significant effect on the torque response of the scissors.

Here, we define the curve of the edge of the upper blade in the Cartesian frame as:

θxy , (2.2.1)

where, (x, y) is a point on the edge of the blade and )(. is a nonlinear function.

We obtain )(. by fitting an analytical curve to the edge of the upper blade. We extract the blade

edge from a real image of the blade.

Considering (2.2.1), the displacement length δ caused by a blade with curve )(. is obtained by:

,h/

,h/

c

c

x2

x2 (2.2.2)

The normal force, fn is calculated by a nonlinear function:

)( gfn (2.2.3)

where, g(·) is a nonlinear function of the point displacement, obtained by measurement of material

properties.

The torque τ caused by fn at the shaft-pivot is calculated by:

Page 57: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

57

cosnc fxτ (2.2.4)

where is the angle between the blade‟s edge and the centerline of the blade. The angle is not

zero because scissors‟ blades are slightly conical as shown in Fig.2.2.2.

The force felt by the user at the handle is calculated by:

cosnc

u fx

FR

(2.2.5)

where R is the distance between the pivot and the handle.

The analytical model accurately predicts the normal applied force.

Force/position control with direct visual observation and force feedback

A pair of scissors is essentially a small robot with two Degrees of Freedom (DOF): rotational

(cutting) DOF and translational DOF. Degrees of Freedom can refer both to how a device keeps

track of position and how a device outputs forces (Fig.2.2.5).

A haptic scissors system is a 2 DOF device with force feedback. A haptic scissors system keeps

track of position with her translational DOF and rotates in the forward-backward by her rotational

DOF and gives forces in those same DOF.

Fig.2.2.5: Two-degree-of-freedom haptic scissors.

Consider a scissors-robot in contact with a rigid object. During the interaction, in the contact

point on the object surface, the measure of the force at the scissors-robot blades can be used to

compute the feed-back torque at the fingers of the user.

The forces of contact between the blade of the scissors and the plate are expressed by a nonlinear

function of deflection at the crack front multiplied by the length of the contact region between

Page 58: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

58

the blade and the plate. On the other hand, the force can be expressed as a function of the object

pose and of the robot position.

The position and orientation of a frame Oo{xo, yo, zo} attached to a rigid object with respect to a

base coordinate frame O{x ,y, z} can be expressed in terms of the coordinate vector of the origin

Oo and of the rotation matrix Ro( o), where o is a ( p × 1) vector corresponding to a suitable

parameterization of the orientation.

In the case that a minimal representation of the orientation is adopted, e.g., Euler angles, it is p = 3,

while it is p = 4 if unit quaternion are used. Hence, the (m × 1) vector Xo = [ OT

o T

o ]T defines a

representation of the object pose with respect to the base frame in terms of m = 3 + p parameters.

The homogeneous coordinate vector p~ = [pT 1]

T of a point P of the object with respect to the base

frame can be computed as p~ = Ho(xo) p~0 , where p~0 is the homogeneous coordinate vector of P

with respect to the object frame and Ho is the homogeneous transformation matrix representing the

pose of the object frame referred to the base frame:

1

000

00 T0

ORxH

)()( (2.2.6)

where 0 is the (3 × 1) null vector.

It is assumed that the geometry of the object is known and that the interaction involves a portion of

the external surface which satisfies a twice differentiable scalar equation 00 )( p .

When in contact, the tip point Pq of the elementary scissors robot instantaneously coincides with a

point P of the object, so that the tip position 0pq satisfies the constraint equation:

00 )( qp (2.2.7)

The unit vector normal 0n to the surface at the point P is expressed in the object frame. The (3 × 1)

contact force vector, of is aligned to the normal unit vector

on to the surface of the object.

Control strategy

This work focuses on techniques for augmenting the performances of haptic scissors system by

means of control architecture.

Haptic scissors capabilities in close interaction with human operator can be considered as

mimicking sensing and actuation of human. This leads to consider fully integrated human vision

and force feed-back based control.

Thanks to the visual perception, the user integrated into a robotic system may achieve global

information on the surrounding environment that can be used for task planning and accidents

avoidance. Human vision and force/torque feed-back are two complementary sensing capabilities

Page 59: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

59

that can be exploited in a synergic way to enhance the precision in the control of a scissors-robot

during the interaction with the environment.

On the other hand, the perception of the force applied to the robot allows adjusting the motion so

that the local constraints imposed by the environment during the interaction are satisfied.

The precision of the accomplished task and dependability of a robotic system are strictly

connected to the availability of sensing information on the external environment. Moreover,

human vision may substitute the complex infrastructure needed for “intelligent environments” to

detect the interaction in the operational space.

Interaction control

When the scissors-robot moves in free workspace, the unknown object pose and the position of the

hand of a human user can be estimated online by using the data provided by the human user eyes.

When the scissors-robot is in contact to the object, also the force measurements and the joint

position measurements are used. Joint values are used for evaluating the position of the scissors for

the interaction configuration with the object.

The Cartesian desired velocity (or force) for the contact (control) points, is transformed in the

corresponding joint trajectory via proper inverse kinematics. Any point on the object can be

considered as a control point. Consider a scissors-robot in contact with an object in the scissors-to-

hand configuration. In the following, some modeling assumption concerning the environment and

the scissors-robot are illustrated.

The proposed algorithm can be used to estimate online the pose of an object in the workspace; this

allows compute the surface equation with respect to the base frame in the form:

000 ),())(( tqopR q

T (2.2.8)

where, the last equality holds in view of the direct kinematics equation of the robot. In the

following, it is assumed that the object does not move; the general case of moving object is more

complex but can be analyzed in a similar way.

Hence, the constraint equation is 0 ),( tq ; moreover 0 qqJ )( , where qqJ /)( is a

Jacobian matrix.

The dynamic model of the manipulator in contact with the object is:

)(),()( qJqqnqqDT (2.2.9)

where D is the (n x n) symmetric and positive definite inertia matrix, ),( qqn is the (n x 1)

vector taking into account Coriolis, centrifugal, friction and gravity torques, τ is the (n x 1)

vector of the joint torques, and ¸ is the lagrangean multiplier associated to the constraint. In the

case of scissors robot we can assume that Coriolis, centrifugal, friction and gravity torques are

less important and can be neglected.

Page 60: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

60

The on-line computation of the constraint equations can be suitably exploited for interaction

control. In the following, the case of the hybrid force/position control is considered.

Hybrid force/position control of the scissors-robot

An inverse dynamics control law can be adopted, by choosing the control torque τ as:

hqJqqnqDT

r )(),()( (2.2.10)

According to the hybrid force/position control strategy, it is useful to apply the change of

coordinates with intermediate variables: hraa PPrr ,,,, defined as:

))()(( qqJaqJ rrrr 1

TT

Pr a0a (2.2.11)

)()( Pr PPdPPdDrPdP rrKrrKra

where KDr and KPr are suitable feedback gains.

The vector rp allows to specify any scissors-robot position and orientation which satisfies the

constraint. The force vector in the new coordinate r is: TF 0 . The scalar λ is computed

from the measured contact force. Velocity vector r and force vector F are orthogonal, i.e.,

0rFT for any r that satisfy the constraint (such that r F = 0).

A force/position control task can be assigned by specifying the desired force Fd = )( td .

The intermediate variable h is calculated as:

))d)(0 (( d

t

Id kh (2.2.12)

with Ik suitable feedback gain.

On the base of the above relations one can configured an inner/outer structure. The inner loop

implements hybrid control whereas the outer loop computes the estimation of the object pose as

well as the desired force and motion trajectories.

The cutting model is computationally efficient, so it can be used for real-time computations of the

corresponding equations.

Simulation results are presented for the case of the cutting process with a pair of scissors in contact

with a planar surface.

Physical prototype of the scissors-robot

Page 61: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

61

In the physical world the static state of deformation in a plastic object under load generally

depends on the forces that are applied to the boundary of the object. When a rigid tool contacts a

deformable object, the position and orientation of the tool with respect to the object uniquely

defines the forces applied to the object. A physical prototype of a scissors haptic device consists

of a scissors-robot with two DOF and an electrical DC-motor. Force/torque feedback is created

through the electrical motor.

Figure 6 is showing the system overview. Experimental haptic system for cutting samples of

paper confirm the model, and is rendered in a haptic virtual environment.

Fig.2.2.6 Scissors device (bottom- side of the picture) and virtual scissors prototype (up- side of

the picture)

The whole system may achieve global information on the environment using vision of the user.

On the other hand, the perception of the force applied to the hand allows adjusting its motion so

that the local constraints imposed by the environment are satisfied.

CONCLUSIONS

The integration of force and visual control to achieve human-robot interaction has been

discussed. A position-force technique based on visual control made by the user and the force

feedback control has been presented, employing a pose estimation algorithm on the basis of

visual direct observation, force and joint position data. A 2-DOF hybrid force/position control

technique was proposed in this paper. The cutting model is used for the haptic system synthesis

for the interaction between a virtual pair of scissors with a virtual plate via the physical scissors

handled by human user. Simulation results have demonstrated the superior performance of the

proposed hybrid control technique with respect to an algorithm using only human‟s visual

perception for the estimation of the object pose. This haptic rendering technique is

computationally efficient and can be applied to real-time cutting process.

Page 62: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

62

Using and testing the haptic devices is a very educational experience. While we knew

approximately what we wished to achieve, we were able to gather valuable feedback by having

the real haptic device in our hands and feeling the impact of torque and force feedback. We

consider that the desktop based workspace could have a great impact when used as a design aid

in conjunction with a virtual prototyping application.

We studied PHANTOM haptic device and we did some qualitative human perception

experiments and informal surveys and questionnaires in an effort to quantify some of the design

parameters of the FALCON system.

Haptic devices can bee integrated into the Delphi software, which simulates maintenance

analysis in a large scale industrial environment. We have created on some desktop virtual

prototyping system for automotive industry. We targeted maintenance analysis applications in

particular. We think that the 6DOF haptic technology in continuing research will provide much

value to virtual prototyping applications on the desktop.

Companies which are most likely to benefit from this technology are those in the automotive

industries, where the parts are many and large, maintainability is a major concern, and it is

prohibitively expensive to produce physical prototypes. All product design can stand to profit

from such a desktop design aid. In the design of any product, it is always important to ensure that

it can be assembled and disassembled speedily and without problems. Haptic technology was

packaged into major CAD systems and it would speed up the product development cycle and

allow better products to be designed and prototyped in less time and cost.

This study introduces a novel structure utilizing the immersive virtual environment for the

interaction between virtual scissors and virtual objects in cutting processes and haptic rendering.

Haptic scissors facilitates precision cutting tasks (maneuvering in constrained spaces, etc.) via

haptic feedback.

Evaluation experiment for basic performance of the device is carried out, and the device is

proved to be able to generate the computed cutting force/torque. A virtual cutting experiment

was carried out, and it proved that the device is able to provide a feeling of cutting. Poor

performances are often attributable to operator error. Operator‟s hand and dynamic factors (e.g.,

changing environment) may still reduce the interface‟s performances. In this case study, the

cooperative human-computer work problem was solved enforcing a „„human-in-the-loop‟‟

strategy.

In this paper we summarized tactile and visual information type and modality combination. The

cognitive properties of cutting modality influence the compatibility of multimodal combinations.

In multitask environment; the cognitive properties of modality also influence how well

information processing can be time-shared with other concurrent tasks.

Summing up, the results of the experiment reported here provide the evidence of a tactile

perception of the displacement effect of the scissors‟ blades on the physical objects. The

influence of cutting force on the virtual object‟s surface was investigated. A study on resulting

surface borders caused by preferred scissors orientation was also conducted. The result reveals

that increasing the cutting force will decrease surface roughness down to a given point beyond

which the surface roughness value continues to decrease as the feed rate decreases.

Page 63: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

63

The novelty of our approach consists of on-line modeling and of integrating the interface

elements into the virtual environment itself. We also take into account the properties of the

edited entities to provide rules which facilitate the construction of 3D objects. It is expected that

working with haptic scissors would prove especially appropriate for opening new ways regarding

sense of touch applications in relation with the utilization of computer systems.

ACKNOWLEDGEMENTS

We would like to thank Thierry Marie Guerra and Michel Dambrine from LAMIH laboratory -

Valenciennes University, France - for their permission to demonstrate the Delphi software for

virtual prototyping applications and for technical information regarding haptic devices. LAMIH

Laboratory supported our collaboration in the haptic scissors project and helped to move this

project from intentions and suggestions level to functional prototypes, among many other

contributions. LAMIH Laboratory is co-inventor of the portable torque-feedback device. One

prototype was developed at Transilvania University of Braso

Chapter 3 The reciprocal robots collision avoidance

3.1 Collision-avoidance using redundant robots

The present article is intended to be a high-level survey over the strategy to on-line collision-

avoidance research that has been carried out in the past ten years. Due to length restrictions, only

a strategy/ technique is explicitly referred to. More exactly, we have selected 10 papers that we

felt were important and different from each other. Also the novelty of the proposed technique,

the number of the citations and the number of downloads were taken into account in the selection

process.

The interest in creating such an overview is diverse. By selecting the most interesting

approaches, we want to focus the attention to new techniques and methodologies that may be of

high interest to the researchers in the field of collision avoidance. Moreover, this selection can be

a useful indicator of the areas that will constitute the future research trends.

The strategy proposed allows the use of redundant degrees of freedom such that a manipulator

can avoid obstacles while tracking the desired end-effectors trajectory. It is supposed that the

obstacles in the workspace of the manipulator are static.

The strategy is based on the redundant inverse kinematics and leads to the favorable use of the

abilities of redundant robots to avoid the collisions with obstacles. This strategy has the

advantage that the configuration of the manipulator can be influenced by further requirements

such as joint limits, etc. The effectiveness of the proposed strategy is discussed by theoretical

considerations and illustrated by simulation of the motion of the four-joint planar manipulators

between symmetric obstacles. It is shown that the proposed collision-free strategy while tracking

the end-effector trajectory is efficient and practical.

Introduction

Particular attention has been devoted to the study of redundant manipulators in the last years.

Redundancy has been recognized as a major characteristic in performing tasks that require

dexterity comparable to that of the human arm.

Page 64: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

64

While most non-redundant manipulators possess enough degrees-of-freedom (DOFs) to perform

their main task(s), i.e., position and/or orientation tracking, it is known that their limited

manipulability results in a reduction in the workspace due to mechanical limits on joint

articulation and presence of obstacles in the workspace.

Redundant manipulators possess extra DOFs than those required to perform the main task(s).

These additional DOFs can be used to fulfill user-defined additional task(s).

The additional task(s) can be represented as kinematic functions. This not only includes the

kinematic functions which reflect some desirable kinematic characteristics of the manipulator

such as posture control, joint limiting, and obstacle avoidance, but can also be extended to

include dynamic measures of performance by defining kinematic functions as the configuration-

dependent terms in the manipulator dynamic model, e.g., impact force, inertia control, etc.

In order to accomplish a task, an accurate joint motion must be commanded to the manipulator.

For that reason, it is necessary to obtain mathematical relations which allow computing joint-

space variables corresponding to the assigned task-space variables. This is the objective of the

inverse kinematics problem.

In this paper the problem of redundant inverse kinematics is reviewed and obstacles avoid

subtask for exploiting the self motion are defined. Various different authors dealt with the

solution of the redundant inverse kinematics problem.

Cartesian position and orientation for the end-effectors task vector xt can be described as a

function of joint variables vector q of the manipulator:

)f( qx t (3.1.1)

While equation (3.1.1) can be obtained easily with the help of Denawit-Hartenberg operators, the

inverse problem is crucial. In the redundant case it is generally not possible to find an inverse

mapping, f −1

.

As an alternative of constructing an inverse mapping, the problem is often reformulated in the

velocities, utilizing the partial derivation of f(q).

The end-effectors task velocity vector tx is:

qqJx )(tt (3.1.2)

with Jacobian matrix:

qqfJ /)(t (3.1.3)

Due to the fact that the inverse of the non-square (analytical) Jacobian Jt(q) does not exist in the

redundant case, the well known generalized Moore–Penrose pseudo inverse Jt+(q)

is utilized.

This proposed strategy often employs a special solution of equation (3.1.2).

Optimization criteria for the redundant self motion can be supplementary by e.g. null-space

projection, which leads to the relation:

0tttt )( qJJIxJq (3.1.4)

Here, )( tt JJI

represents the orthogonal projection matrix in the null space of Jt, and 0q

is an arbitrary joint-space velocity; the second part of the solution is therefore a null-space

Page 65: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

65

velocity. The vector 0q can be chosen arbitrary and is used to force a desired behavior for the

null space motion. The null space of Jt (q) is defined as:

})(0{))(( tt qqJ:QqqJN

(3.1.5)

In redundant directions joint velocities causes no motion at the end-effectors level. These are

internal motions of the manipulator. Redundant joint velocities satisfy the equation:

0t q(q)J (3.1.6)

At each configuration, the null space of Jt is the set of joint-space velocities which yield zero

task velocity; these are thus called null-space velocities.

AUGMENTED JACOBIAN

The redundant robots can operate in the Cartesian space with obstacles. The redundant self

motion can satisfy both, the end-effector task and the additional constraint task cause of the

obstacles, at the same time.

The concept of task-space augmentation introduces a constraint task to be fulfilled along with the

end-effector task. In that case, an augmented Jacobian matrix is set-up whose inverse gives the

required joint velocity solution.

Let us consider the p-dimensional vector xc = (xc,1 . . . xc,p) which describes the additional tasks

to be fulfilled in addition the m-dimensional end-effector task vector xt, i. e., the degree of

redundancy p = n – m , whereas n is the number of joints.

The relation between the joint-space coordinate vector q and the constraint-task vector xc can be

considered as a direct kinematics equation:

)(cc qkx (3.1.7)

where kc is a continuous nonlinear vector function. By differentiating (3.1.7) one can be obtained

(3.1.8):

qqJx )(cc . (3.1.8)

In (3.1.8) cx is the constraint-task velocity vector, and the mapping Jc(q) = ∂kc/∂q is the (p x n)

constraint-task Jacobian matrix. At this moment, an augmented-task vector ax , can be defined by

stacking the end-effector task vector with the constraint-task vector as:

Tct

T

cta )()( qkqkxxx . (3.1.9)

According to this definition, finding a joint configuration q that result, in some desired value for

ax , means satisfying both the end-effector task and the constraint task, at the same time.

At the differential level, Cartesian velocities are given by:

qqJx )(aa (3.1.10)

To finding a joint velocity q in some desired value for Cartesian velocities ax can be used:

Page 66: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

66

0aaaa )( qJJIxJq (3.1.11)

That means the inverting of the augmented Jacobian, Tcta )()()( qJqJqJ to obtain the pseudo-

inverse matrix, Ja+.

Task priority strategy

With reference to solution (3.1.11), the task-priority strategy consists of computing 0q so as to

correctly achieve the p-dimensional constraint-task velocity, cx .

In the typical case, an end-effector task is considered as the primary task, even if examples may

be given in which it becomes the secondary task. The idea is that, when an exact solution does

not exist, the reconstruction error should only affect the lower-priority task.

Conflicts between the end-effector task and the constraint task are handled in the framework of

the task-priority strategy by correctly assigning an order of priority to the each given tasks and

then satisfying the lower priority task only in the null space of the higher-priority task.

Remarkably, the projection of 0q onto the null space of Jt ensures lower priority of the constraint

task with respect to the end-effector task because this results in a null-space velocity for the

higher-priority task.

When the secondary task cx is orthogonal to the primary task tx the joint velocity:

cc0 )( xqJq (3.1.12)

would easily solve the problem, being in addition already a null-space velocity for the primary-

task velocity mapping (3.1.2).

However, in the general case the two tasks may be compatible but not orthogonal and there not

exist a joint velocity solution that ensures the achievement of both tx and cx . Coherently with

the defined order of priority between the two tasks, a reasonable choice is then to guarantee exact

tracking of the primary-task velocity, while minimizing the constraint-task velocity

reconstruction error qqJxε )(cc ; this gives:

ttccttc0 )( xJJxJJIJq (3.1.13)

The solution given by (3.1.4) and (3.1.13) can be simplified to:

ttccttctt xJJxJJIJxJq )( (3.1.14)

Another approach is to relax minimization of the secondary-task velocity reconstruction

constraint and simply pursue tracking of the components of (12) that do not conflict with the

primary task, namely

cctttt xJJJIxJq

)( (3.1.15)

An intuitive justification of this solution can be given as follows: The pseudo inverses Jt+ and Jc

+

are used to solve separately for the joint velocities associated with the respective task velocities.

The joint velocity associated with the (secondary) constraint task is then projected onto the null

space of Jt to remove the components that would interfere with the (primary) end-effector task.

Page 67: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

67

Finally the joint velocity associated with the constraint task is added to the joint velocity

associated with the end-effector task.

By construction, the solution (3.1.15) leads to larger constraint-task reconstruction errors than

solution (3.1.14); this is the price paid to give smooth and feasible trajectories for the joint

velocity in tracking conflicting tasks.

Physical constraint problem Physical constraints, like joint speed limits can be enforced in a natural way via inequality

constraints. Additional constraints in form of inequalities can be added to restrict the solutions

for particular joints (e.g. restrictions due to hardware constraints or maximum joint movements

per time step, Ta ).

maxmin , qqq (3.1.16)

Using the following incremental formulation:

aTqq (3.1.17)

aTxx (3.1.18)

for the null-space motion (3.1.6), linear conditions can be chosen of the form:

aqA (3.1.19)

So the inverse kinematics problem can be formulated as an optimization problem

)(min IKq

qL

(3.1.20)

for the cost function

2IK )( aqAqL (3.1.21)

with the equality constraint as main task

qJx tt (3.1.22)

Herein the symbol2

. denotes the quadratic norm. Notice that the main task forms in fact the

constraint for a linear optimization problem here.

The particular equations:

ii aqA (3.1.23)

are weighted by the factors wi, which define how much an error in the ith

equation affects the

value of the cost function LIK(Δq).

It is now possible to combine multiple behaviors for the null-space motion by choosing

Tjj11 ww AAA ........ ; Tjj11 awaw .....a (3.1.24)

Page 68: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

68

It should also be noted that this optimization problem leads to the pseudo inverse solution by

choosing A as the identity matrix and a as a vector containing just zero elements.

Collinsion avoidance problem The null-space motion concept is use for the avoidance of collisions, of the robot arm, with an

external object. Collision avoidance is here only presented for pure null-space movements. Only

collisions of the robot‟s internal structure with the” external object”, is represented. For that

reason an artificial potential field is attached to the considered external object.

Let d denote the minimal distance between the robot and the external object. If this distance is

smaller than a proper chosen value d0 defined in the collision area, then the robot should move

away according to the potential field V (d).

2

0 )(2

1)( dddV (3.1.25)

For the null-space motion an equation of the form (3.1.13) is needed. Assume, that the robot

moves sufficiently slowly, then the change of the value of V(d) due to an joint movement can be

formulated as follows:

qq

d

d

V(d)V(d) (3.1.26)

Equation (3.1.26) can now be transposed as the (3.1.23) form, with

q

A

d

d

V(d)1

(3.1.27)

It is obvious that collisions of the end-effector with the collision-object cannot be handled in this

way, because only self motion has been considered herein. To avoid collisions of the end-

effector too, it is necessary to allow a deviation of the desired trajectory, which can be calculated

similar to (3.1.27).

For interaction between the user and the collision system a position tracker is utilized.

For every obstacle a boundary ellipse is defined in workspace such that there is no collision if the

robot joints or end-effectors are outside these ellipses. The collision-avoidance algorithm

computes on-line the self-motion movements as function of distance d. The movements of a

reference frame, which is fixed to a particular joint, can be commanded the same as it don‟t

affect the trajectory of the end-effector.

A. Delphi simulators for validation of the collision-avoidance strategy

The authors propose the Delphi informatics environment for the qualitative simulation of the

robotic systems, using the visual programming. The prototype graphical simulation system has

been developed to support and demonstrate different aspects of robots behavior.

In traditional robotics, programmers attempted to anticipate and explicitly control every aspect of

the action of the robot. However, these systems often failed when are placed in a changing

environment, where unanticipated situations appear.

Systems using autonomous agents, with de-centralized control of robot motion, enable robots to

learn and adapt to changing environments. When in the environment appear accidentally the

Page 69: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

69

unanticipated objects, the robot control de-centralized system must be able to select the optimal

path for avoid the collision between robot arm and strangers objects.

Autonomous agents are software programs designed to pursue goals and can do so without user

intervention. The idea is that agents are not strictly invoked for a task, but activate themselves.

They are usually large applications, which attempt to solve practical problems demanded by the

additional tasks. Advanced robotic systems involve large agent-based simulators aimed at

understanding "action selection". By this mechanism the robot selects the behavior to execute

from many possible behaviors.

Delphi simulators are generally smaller software programs and they are written in order to

evaluate and validate by simulation the robots‟ behavior. They usually involve algorithms that

allow exposing the behavior of the virtual robots and their virtual environment. This technique is

used also to provide the command methods for the real robotic systems.

All of these programs enable the manifestation of global behaviors. They also attempt to

simulate more aspects of robot behavior - not just movement. As such they include the

movement of robot manipulators, the contact with manipulated objects and the collision with

unexpected objects into the work environment.

All components of the artificial world are objects, as is the environment itself. The interface

consists of a collection of tools for analysis and visualization.

Data analysis objects can communicate with the user interface objects to present displays of data,

as well as for storage to files. The prototype allows the user to create, modify, and destroy

objects. Each object has a few standard attributes managed by Object Inspector, as well as user-

specified private data [6].

In this paper the graphical simulation prototype system has been developed to support and

demonstrate different aspects of robots behavior when met the obstacles in his work

environment. All of these programs are write in Delphi language and enable the manifestation of

global behaviors.

For itself they include the movement of virtual robot manipulator, the contact with virtual objects

and the collision avoid, into the work environment, with expected or unexpected virtual objects.

The information regarding the collision are given about the shape of each face of the body and

then it figures out which bodies face touch each other and passes the resulting contact point

visual information to designer. The designer can then take the proper decisions concerning the

real prototype.

To implement a real-time collision-avoidance strategy, robot and environment modeling and

distance calculation need to be investigated. An obstacle-avoidance system essentially deals with

a complex environment.

There are many limitations in creating (modeling) a robot‟s environment such as space and

equipment limitations. Obviously, the accuracy with which a robot arm and its environment are

modeled is directly related to the real-time control requirements.

Greater detail in modeling results in higher complexity, when computing the critical distances

between an obstacle and the manipulator is needed.

Much of this computation can be avoided if the distance measurements are obtained by a

proximity sensing system. For situations where proximity sensors are not available, a possible

solution is to use simple geometric primitives to represent the virtual arm and its virtual

environment.

A simplified geometrical model for links of industrial manipulators is relevant to the study of

collisions either with each other or with objects in the workspace.

Page 70: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

70

In this paper the joints were represented by circles and the links were modeled by straight lines

and rectangles.

A redundancy-resolution scenario was proposed to achieve obstacle avoidance. In the following

simulations the main task consists of tracking the position and orientation trajectories, generated

by linear interpolation, between the initial and final poses. It should be noted that interpolation of

rotations is a much more complex problem than point interpolation [7]. For this reason, we use

simple linear interpolation for both translation and rotation, which nevertheless leads to

satisfactory results.

In the simulation presented in this section the redundancy resolution is implemented in closed-

loop. One solution to this problem is online transmission of a robot configuration to a

workstation running a graphics visualization of the arm how serves as a virtual environment; the

graphics model of the robot mirrors the exact motion of the arm, and the environment can be

modeled in the graphics program. This approach has two main advantages:

• Any complex environment can be modeled with

a desired precision (including a time-varying environment)

• The risk of damage to the robot is reduced.

Few different scenarios were selected to verify the performance of the obstacle-avoidance based

redundancy-resolution strategy in executing the following tasks: position tracking, orientation

tracking, stationary symmetric obstacles collision avoidance, joint limit, and self-collision

avoidance. In each of these scenarios, one or multiple features were active at different instants of

execution.

The simulation methodology uses possible scenarios and involves some sequences of animation,

as one can see in the Fig.1, Fig.2 and Fig. 3 for a virtual assembly system with a planar

manipulator. In this simulation, the end-effector is initially at rest. At the end of the free path the

end-effector touches the bolt at the contact point with the head of the bolt.

The robot arm pushes the bolts in the symmetrical holes until the bolts touches constraint surface

at the base of holes. After the contact between the end-effector and the bolt the success of the

main task consists of keeping the position of the end-effector in the horizontal direction, while

exerting a desired force in the same direction.

There is also the support object enclosed in area collision ellipse, in the workspace. The

additional task consists of using the redundant degree of freedom to avoid the horizontal and

vertical baseline of this support objects. The manipulator moves free until the arm is close

enough to the support object. The obstacle avoidance task becomes active and makes the

manipulator move in the null space of the Jacobian matrix to avoid collision.

The collision avoidance is here only presented for pure null-space movements. So only collisions

of the virtual robot‟s structure with the ”virtual collision object” are handled.

B Results of testing in virtual environment

The figures below show the results of the simulations for different poses of the robot arm

corresponding to the collision avoidance task. In a practical implementation, no collision occurs

if the maximum acceleration of each joint would be limited and this commanded joint

acceleration would result in saturation of the actuators.

Page 71: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

71

Fig. 3.1.1 Beginning pose of the animation sequence

Fig.3.1.2 Middle pose of the animation sequence

area collision area collision

area collision

area collision

Page 72: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

72

Fig. 3.1.3 Ending pose of the animation sequence

From preliminary simulations, it was observed that finding a maximum acceleration value which

performs well in different situations is very difficult. To overcome this problem, a time-varying

formulation must been used to adjust the maximum acceleration automatically.

We tested virtual architectures the same way as the reinforcement strategy. Virtual robot arm

moves many times consecutively.

We consider as distance imposed in the constraint space the distance of end-effector from the

environment horizontal baseline reference, in the X-axis and for vertical baseline, in Y-axis

respectively. If this distance is between the 20 pixels from the horizontal baseline and vertical

baseline respectively, we considered the trial as success. On the other hand, if end-effector

position X -axis and Y-axis respectively goes away more than 20 pixels from the baseline at any

move, we considered it as failure.

We have demonstrated the validity of the strategy using a scenario where linear constraints are

imposed at lower priority task.

Conclusion

In this paper, we have considered a redundant inverse kinematics model based on the augmented

Jacobian that enables the use of self motion of the manipulator to perform as sub-task the

obstacles avoidance.

According to the task splitting concept in (primary) end-effector task and (secondary) constraint

task, the concept was demonstrated on a virtual system. Interaction between the virtual robot and

virtual obstacle was demonstrated, while performing end-effector tasks on the robot application.

The virtual robot is able to react to avoid the obstacles while executing the primary task. This

was shown experimentally with a virtual system. This kind of robot‟s behavior simulation will be

necessary in real applications, where interactions between robots and his environment are vital.

The performance of redundancy resolution in tracking the main task trajectories is studied here

by computer simulation.

We can apply an command strategy from virtual system to real system without any complicated

control theory or position sensors. Moreover, if we use different shaped obstacles it‟s very

difficult to find proper control rules, but we can use the virtual robot learning by applying this

strategy to get the satisfactory result with the real robot arm.

area collision

area collision

Page 73: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

73

In this way, the kinematic simulation running on an workstation, generates the desired joint

trajectory and this trajectory is then transferred as the joint set points to the robot‟s joints PID

controllers. The real robot joints movement will imitate the virtual robot joints movement.

We wrote a program that allow to the end-effector of the virtual robot to wander around without

bumping into obstacles in its path. The performance of the obstacle avoidance simulation

strategy has been studied by various simulations for different scenarios. The performance of the

collision avoidance algorithm was observed using the virtual models of the arm and of the object

in a virtual environment.

In this paper, we implemented a very simple program that worked effectively. In the virtual

environment not all simulation programs need to be complex to be effective. It is easy to imagine

that, in a practical implementation, the software algorithms for real robot are highly complex.

A redundancy-resolution scenario was proposed to achieve obstacle avoidance. In the following

simulations the main task consists of tracking the position and orientation trajectories, generated

by linear interpolation, between the initial and final poses. It should be noted that interpolation of

rotations is a much more complex problem than point interpolation [7]. For this reason, we use

simple linear interpolation for both translation and rotation, which nevertheless leads to

satisfactory results.

In the simulation presented in this section the redundancy resolution is implemented in closed-

loop. One solution to this problem is online transmission of a robot configuration to a

workstation running a graphics visualization of the arm how serves as a virtual environment; the

graphics model of the robot mirrors the exact motion of the arm, and the environment can be

modeled in the graphics program.

Page 74: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

74

3.2 On-line collision avoidance using Programming by Demonstration

In this paper, based on original idea, the author proposes a new strategy to robot collision

avoidance using programming imitation paradigm. To program the desired motion sequence for

the physical robot, one captures the motion reference paths from her virtual robot model and

maps these to the joint settings of the physical robot. Motion imitation requires transfer of a

dynamical signature of a movement of the virtual robot to the physical robot, i.e. the robots

should be able to encode and reproduce a particular path as one with a specific velocity and/or an

acceleration profile. Furthermore, the virtual robot must cover all possible contexts - including la

presence of accidental obstacles - in which the physical robot will need to generate similar

motions in unseen context.

INTRODUCTION

Programming by Demonstration has appeared as one way to respond to growing need for

intuitive control methods and is one of the most promising programming techniques for robotic

manipulators. This technique allows even inexperienced users to easily program the robots based

on the teaching by imitation paradigm.

This work follows a recent trend in Programming by Imitation. We present a new method to

programming physical robot through the imitation of the virtual robot prototype.

Movement imitation requires a demonstrator. In our approach the demonstrator is the virtual

robot prototype and the imitator is the physical robotic manipulator.

Conformity our method the dynamics of the motion of the virtual robots is reproduced by the

physical robot.

In this paper one use the virtual robot prototypes and the motion capture systems to obtain the

reference motion data, which typically consist of a set of trajectories in the Cartesian space.

We restrict our study to imitation of manipulative tasks. In particular, the ability to imitate virtual

robots gestures and pursue task-relevant paths is essential skills for physical robotic

manipulators, in the spirit of our method.

Inspired by the proposal patent [11] and based on the motion imitation concept this paper

develops a general policy for robot motion programming based on virtual robot prototypes.

The strong point of the proposed method is that it provides fast on-line re-planning of the motion

in the face of spatial-temporal perturbations as accidental obstacles and it allows generalizing a

motion to unseen context based on particular scenarios.

This approach is focused on tracking joint angle trajectories, while some tasks may require

tracking other quantities, such as end-effectors trajectories which will be addressed in future

work. The objective of the present paper is giving an overview of a new robot programming

method based on the virtual prototypes.

OVERVIEW OF PROGRAMMING BY IMITATION BASED ON NEW STRATEGY

Imitating the motion with stable robot dynamics is a challenging research problem [5].

Conventional programming strategies are hardly applicable to robotic manipulators that joints

must be coordinated concurrently and must assure a stable dynamics [10]. Therefore, exploiting

virtual robots‟ potential in programming tasks remains an atypical objective.

A characteristic feature of robot programming is that usually it is dealing with two different

worlds; the real physical world to be manipulated and the abstract models - in particular virtual

Page 75: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

75

prototypes models - representing this world in a functional or descriptive manner by programs

and data. In the simplest case, these models are pure imagination of the programmers; in high

level programming languages, e.g. it may consist of CAD data.

Creating accurate robot path points for a robot application is an important programming task. It

requires a robot programmer to have the knowledge of the robot‟s reference frames, positions,

operations, and the work space.

In the conventional “lead-through” method, the robot programmer uses the robot teach pendant

accessory to position and to orientation the robot joints and end-effector and record the satisfied

robot pose.

The basic idea behind these approaches is to relieve the programmer from knowing all specific

robot details and liberate him from coding every small motion.

Today‟s robot simulation software provides the robot programmer with the functions of creating

virtual robot and virtual path points in an interactive virtual 3D design environment.

By the time a robot simulation design is completed; the simulation robot program is able to move

the virtual robot end-effector to all desired virtual path points for performing the specified

operation. By different scenarios one creates particular motion paths in the virtual space that are

selected for particular motion paths in physical work space.

However, because of the inevitable dimensional differences of the components between the

physical robot work space and the simulated robot work space, the virtual robot path points must

be adjusted relative to the actual position of the components in the physical robot work space.

This task involves the techniques of calibrating the position coordinates of the simulation device

models with respect to the physical robot path points.

Learning by imitation based on our method represents a new research topic in robotics being a

promising approach, towards effective robot programming.

FORMULATION OF THE IMITATION TASK

An imitation task can be decomposed into the serial implementation of two processes: an

observation process and an imitation process. The observation process consists of extracting

relevant (i.e. in our case, invariant over time) features from a demonstrated dataset.

The imitation process consists of generating an imitated dataset that minimizes the discrepancy

between the demonstrated and imitated datasets. Formalism Let S be the dataset generated by the demonstrator while driven by a control strategy U. The

control strategy U is such that },{)( DIUS , where },,{ DDDD (angular position, speed

and acceleration) the trajectory of the demonstrator‟s arm joints and },,{ IIII (angular

position, speed and acceleration) the trajectory of the imitator‟s arm joints.

The imitation process consists of determining a control strategy U’, that generates a dataset

}','{)'(' DIUS such that the cost function J, or the metric of the imitation task, is minimal:

0)',( SSJ .

Each imitation task is defined by a set of constraints C with {c є [1, C]}. For each constraint c, є

a control strategy Uc, generating a dataset S U c, such that the associated metric J U c is minimal

[13]:

0)',( UcUcUc SSJ (3.2.1)

Page 76: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

76

Metric of imitation

We express the metric or cost function of the imitation task J, as a linear combination of

constraint-dependent cost functions JUc :

)',()',(1 1

'

U

u

C

ccUUcUc SSJwSSJ (3.2.2)

The weights of the constraints-dependent cost functions are proportional to the probability P

(Us), that the dataset S has been generated by the control strategy Uc:

Ui

Cj j

UciP

UcPw

1 1 )(

)( (2.3.3)

The weights are normalized, such that:

1)(0 1

U

i

C

j

Ucw (3.2.4)

Optimal imitation policy

We hypothesize that the constraints c, once identified during the observation process, remain the

same during the imitation process. In the particular case where all control strategies are mutually

exclusive, the optimal control strategy is unique [9].

If P (U’c) is the probability that the imitated dataset has been generated by the control strategy

U’c, we have:

P(Uc) = P’(U’c) (3.2.5)

In the case where different control strategies can coexist, because they would be acting on

separate sub-datasets (e.g. one strategy tends to minimize a constraint on the objects trajectories,

while another strategy tends to minimize a constraint on the arm joint trajectories), then the

optimal control strategy is the combination of all the optimal control strategies for each sub-

dataset, such that:

P (UC) = P’ (U’C) ∀ c є [1, C]. (3.2.6)

This probabilistic framework is highly suitable to the analysis and production joint trajectories,

when using probabilistic methods.

Metric of size

Movement imitation requires a demonstrator. In our approach the demonstrator is the virtual

robot prototype and the imitator is the physical real robot. The inputs of the imitation procedure

are virtual robot prototype behavior captured, as virtual joint trajectories represented such a

sequence of angular values of its joints. The imitator imitates the original virtual prototype

captured motion, as closely as possible, and at the same time this respects they physical limits.

The virtual robot prototype behavior‟ problem will be studied by computer graphics, and we

have actually much commercial graphic software that can solve it efficiently.

Generally, the virtual robot prototype is modeled in computer graphics with an approximation,

and in addition its links lengths are largely different from those of physical real robot. Its size

must be scaled to fit the physical robot size. However, in the motion imitation by a physical real

Page 77: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

77

robot, additional difficulties arise such as the joints variables, angular velocity and torques limits

[4].

By taking into account those additional constraints, the motion imitation problem becomes the

dynamic equation of physical robot arm:

rrrr qqNqqM ),()( (3.2.7)

The right term τr is the vector of the applied torques on the physical real robot‟s joints.

Permanent constraints are followings:

or qq 0 , 00rq , 00 rq

frf qq , 0frq , 0rfq (3.2.8)

qqq r ,

qqq r ,

r

_,

where q−, q

+ and denote the minimal and maximal values of vector qr and vector τr respectively.

Solving the optimization problem, in the case of above constraints, is generally difficult.

The explanation of the involvement of the dynamic equation of motion and the desired virtual

path points for performing the specified operation becomes implicit relation between the vector

of applied torques τr and the vector of joints positions qr of the physical robot.

PROGRAMMING FROM A SIMULATED VIRTUAL MODEL

We present a description of the theoretical aspects of the programming from a simulated virtual

model and of the strategy which is at foundation of this approach. The advantages of such

programming approach as an alternative to the classical methods (e.g. vision guided trajectory

imitation) are on-line adaptation to the motion of the virtual prototype.

A solution to the above problem is to construct a virtual prototype model and transfer the virtual

trajectory by interacting with the physical model.

Designing a model would be an option; however, the behavior of the robots is very difficult to

model. Moreover, the use of system knowledge is contrary to our research aim. Therefore we

focus on creating a virtual prototype model from experimental data obtained from the physical

robot model.

Thus, we will prove that our method guarantees the motion optimization for each robotics tasks.

The planning package communicates primarily with simulation environment. A planning module

can send messages to the simulation system such as computed plans for the robots. The planning

module can further send trajectory and planning structure information to visualization so users

can see the results of an algorithm.

The planning module also receives control signals from the simulation module, such as when to

start planning joint trajectories [2].

The visualization module is in charge for visualizing any aspect needed by the programmer for

the optimization process. Users interact with the simulation environment through the

visualization. This includes, but not limited to, computer screen. Optimization of the real robots

behavior is performed in the low dimensional virtual space using the virtual robot prototypes. In

the virtual space one simulate even the intersecting of the virtual robot and her environment. The

intersecting of two virtual objects is possible in the virtual world, where the virtual objects can

be even intersected and there no exist the risk to be destroyed.

Page 78: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

78

The contact detection between the virtual objects is responsible for determining the distance

between two points in a space and for optimization of the trajectories of motion. This strategy

may use simple CAD methods or may be extended to be more complex and take into account

invalid areas of the space. The visualization provides an interface to develop interactive

implementations based on imitation strategy.

In our work, we assume that learning of the deterministic part for description motion dynamics

should be sufficient to design the corresponding robot control.

We particularly refer to the ability of the system to react to changes in the environment that are

reflected by motion parameters, such as a desired target position and motion duration. Therefore,

the system is able to manage with uncertainties in the position of a manipulated object, duration

of motion, and structure limitation (e.g., joint velocity and torque limits).

The proposed method aims at adapting to spatial and temporal perturbations which are

externally-generated. This aspect will be investigated in our future works.

ONLINE IMITATION OF THE DYNAMICAL SYSTEMS

In robotics, one of the most frequent methods to represent movement strategy is by means of the

learning from imitation. Imitation learning is simply an application of supervised learning. One

goal of imitation of the dynamical systems is to use the ability of coupling phenomena to

description for complex behavior [6].

In this paper, we propose a generic modeling approach to generate virtual robot prototype

behavior in experimental scenery. The actions for the each task are computed for virtual robot

prototype and are transferred online, with a central coordination, to corresponding physical

robot, which must imitate her virtual “homonymous”. Notice the similarity between moves of the

virtual robot prototype in the virtual work space and the “homonymous” moves in the real work

space of the physical robot. We assume to use the virtual robot prototypes and the motion

capture systems to obtain the reference motion data, which typically consist of a set of

trajectories in the Cartesian space.

The paper relates to a method and a robot programming platform by combining off-line and on-

line programming techniques. The method consists in using a programming platform on which

there is carried out the virtual prototype of the physical robotic arm to be programmed and the

real working space wherein it is intended to work.

In the robot program there is written a source code intended to generate the motion paths of the

virtual robotic arm prototype. The numerical values of the prototype articulation variables are

sent to the data register of a port of the information system which, via a numerical interface, are

on-line transferred into the data registers of the controllers of the actuator of the physical robotic

arm. Finally, there are obtained tracking structures due to which the moving paths of the virtual

robotic arm joints are reproduced by the physical robotic arm joints, thereby generating motion

within the real working space.

Imitation learning from our strategy demonstrates how to obtain dynamical virtual models with

CAD systems. Those online adjusted virtual models are among the most important properties

offered by a dynamical systems approach, and these properties cannot easily be replicated

without the feed-back from physical robot of our proposed structure.

Page 79: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

79

The objective of a movement is to generate a reaching movement from any start state to a goal

state [8]. The discrete dynamical system is initialized with a minimum movement, which is

frequently used as an approximate model of smooth movement.

The proposed structure uses a virtual demonstrator for planning the movements of the physical

robot to avoid the obstacles. We investigate the potential of imitation in programming robotic

manipulators with multiples degrees of freedom when the associated joints must be coordinated

concurrently.

The imitation strategy consists in a proportional real-time mapping between each virtual joint

and the corresponding physical joint. The system requires the programmer to perform an initial

calibration routine to identify the range of motion for each movement. Each range is divided into

as many intervals as the number of feasible discrete configurations of the corresponding joint.

EXPERIMENTAL CONFIGURATION FOR ROBOT PROGRAMMING BY IMITATION

The approach was implemented using a simulation software platform called Robot Motion

Imitation Platform (RMIP) developed at University of Brasov [11], who its feasibility has been

demonstrated in some simulation settings for manipulative works, using Virtual Reality.

We extended the simulation software platform to programming a physical robot. We see them as

intelligent structure that can be used to transfer - in real time - the gesture of the virtual robot to

the physical robots for generating complex movements in real work space. One transfers via

intelligent interface, the virtual joint angles data from a motion capture system to a kinematic

model for a physical robotic manipulator.

The RMIP is an architecture that provides libraries and tools to help software developers in

programming. Platform RMIP is focused on 3D simulation of the dynamics systems and on a

control and planning interface that provides primitives for motion planning by imitation.

Also is a object-oriented infrastructure for the integration of controllers, as well as the integration

of planners with controllers to achieve feedback based planning, In particular, RMIP is used to

provide concrete implementations of motion planners.

The platform RMIP utilizes a framework which allows both high level and low level controllers

to be composed in a way that provides complex interactions between virtual and physical robots.

In our strategy we just included the possibility of functional coupling virtual and physical robot

model. The functional coupling consists to map of the motion of the demonstrator robot arm with

the imitator robot arm. The former is displayed on the screen, giving the visual information to

programmer and is connected to the latter through intelligent interface. In our experiments we

have adapted as imitator a small robotic manipulator, named EXMAN and presented in Fig.

3.2.1.

Fig. 3.2.1 Experienced robotic manipulator

Page 80: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

80

We start with a 3 degree-of-freedom (DOF) discrete movement system that models point-to-

point attained in a 3D Cartesian space.

The arm moved in a free space without obstacles, had three active joints (shoulder, elbow and

wrist) and was driven by electrical actuators.

Figure 3.2.2 shows our experiment involving the imitation learning for a physical robotic arm

with 3 degrees-of-freedom (DOFs) for performing the manipulate tasks. We demonstrated the

imitation of elbow, shoulder and wrist movements. Importantly, these tasks required the

coordination of 3 DOFs, which was easily accomplished in our approach.

The imitated movement was represented in joint angles of the robot. Indeed, only kinematic

variables are observable in imitation learning. The robot was equipped with a controller (a PD

controller) that could accurately follow the kinematic strategy (i.e., generate the torques

necessary to follow a particular joint angle trajectory, given in terms of desired positions,

velocities, and accelerations) [7].

Fig. 3.2.2 Imitation software platform structure

To generate the motion sequence for the real robot, one captures the motions from a virtual robot

model and maps these to the joint settings of the physical robot. Initially, a set of virtual desired

postures is created to the virtual robotic arm BRV and the pictures‟ positions are recorded for

each posture, during motion.

These recorded pictures‟ positions provide a set of Cartesian points in the 3D capture volume for

each posture. To obtain the physical robot postures, the virtual pictures‟ positions are assigned as

positional constraints on the physical robot. To obtain the physical joint angles one use standard

inverse kinematics (IK) routines. The IK routine then directly generates the physical joint angles

on the physical robot for each posture.

Referring the Fig. 2 we comment the following: on programming platform, a robot program is

carried out off-line, and one sends into the data registers of a port of the hardware structure, the

numerical values of the joint variables of the virtual prototype of the robotic arm (BRV) and

displays on a graphical user interface, the evolution of the virtual prototype during the carrying

out of the robotic task. Via numerical interface (IN) the virtual joint dataset, from the data

registers of the port of the hardware structure of the programming platform are transferred into

the data registers of the numerical comparators of the controllers. These datasets are reference

inputs of the pursue loops, resulting a control system (SC).

The virtual tracking paths of the virtual joints are reproduced in the real work space by the

actuators of the physical robotic arm. The reference datasets are obtained using a motion capture

Page 81: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

81

channel taking into account the joints motion range. The symbolic spatial relations specifying the

virtual work space can be used for the automatic pursuits of possible virtual path as well as for

planning of appropriate behavior of the physical robot arm BRR, which may guide the motion

process during task execution.

The easiest way to generate the spatial relations explicitly is the interactively programming of

the behavior of the virtual prototype in his virtual environment, in order to specify suitable

positions θv1, θv2, θv3. This kind of specification provides an easy to use interactive graphical tool

to define any kind of robot path; the user has to deal only with a limited and manageable amount

of spatial information in a very comfortable manner.

The robot programming system has to recognize the correct robot task type and should map it to

a sequence of robot operations [11]. The desired pathways are automatically transferred and

parameterized in the numerical interface IN, using the path planner.

The applicable robot tasks are designed and the desired pathways are programmed off-line and

stored in the buffer modules RT1, RT2, RT3.The comparative modules CN1, CN2, CN3 furnish,

to the pursuit controllers, the datasets involving the expected state of the virtual robot prototype

and the measured state of the physical robot.

While motion execution is in progress, the real robot joints ARR1, ARR2, ARR3 are activates

into the real work space. Each actuator was connected by a sensor in the closed-loop. Each time,

a skill primitive is executed by the robot control system SC; it changing the robot joints position.

As no time limit for the motion is specified, the physical robot imitates the behavior of the

virtual robot.

In our laboratory currently we are developing Cartesian control architecture able to interpret the

physical robot commands in the above given form. The basis of our implementation is a flexible

and modular system for robot programming by imitation.

In our experimental configuration in order to prove the correctness of the robot programming by

imitation we have chosen an robotic manipulator equipped with electrical actuators, mounted on

the physical robot‟s joints.

The robot‟s control unit is connected via TCP/IP to a PC equipped with the interface card; the

PC is running the simulation and control process. The robot control system receives and executes

each 16 ms, an elementary move operation.

Due to the kinematics limitations of physical robot, the resolution of the joints is quite limited.

After the calibration phase, physical robot starts imitating the gestures of the virtual robot

demonstrator. The user can decide to activate all the three degrees of freedom concurrently or

with a restricted subset by deactivating some of the sensors attached to each joint of the physical

robot. Our system requires an essential step in that one converts the position errors into motor

commands by means of the PD controller [7].

Figure 3 shows the structure of the system used for joint motion imitation. One can see the

interest hardware components their interconnections for closed-loop control, associated of robot

joint number one. The sensor TP1 is in close proximity with the actuator ACT1 and is

interconnected with hardware components specific for a closed-loop.

We aim at developing controllers for learning by imitation with a single joint robot demonstrator.

For this purpose, we assume a simple control system where the position of the 1 DOF discrete

dynamical system, drives the time evolution of joint variable θr1, which can be interpreted as the

Page 82: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

82

position controlled by a proportional-derivative controller. All other DOFs of the arm

demonstrate the same behavior.

Fig. 3.2.3 Closed-loop motion control for robot joint

Torque is indirectly controlled via joint dynamics. The internal joint states were unknown and

not available to the controller. Only joint variable was available by the joint sensor. The joint

kinematics is derived from the CAD model. In the motion imitation additional difficulties arise

such as the joints angular velocity and torques limits.

RESULTS-EVALUATIONS

In this paper, we presented and reviewed several experimental evaluations of applying our

approach in the domain of motion programming based on virtual robots prototypes, using both

simulation and robotic control studies.

The evaluations were intended to demonstrate the properties of our methodology, but also the

domain-specific choices that need to be made. We have discuss about imitation learning of

discrete movement, online imitation with the help of coupling parts, synchronization phenomena,

and movement recognition based on a graphical interface.

Experiments involving gesture imitation were carried out to test the effectiveness of the system.

The evolution of the references signals, sent to the physical robot at level of each degree of

freedom, has been correlated with the resolution of the corresponding virtual “homonymous”

degree of freedom. In consequence the physical robot joints follow with acceptable fidelity

imposed references signals values.

The platform RMIP is based on original idea patented by author. Abstract of the proposal patent

illustrating this method can be viewed at reference [14].

One expected that this method, being user-friendly, would enhance the application of

programming robots in dynamic environments. We anticipate that these promises may be

fulfilled very soon.

The imitation approach illustrated in this paper is rather general, even though its implementation

is constrained by limitations of RMIP. In particular, the paper reports successful experiments on

gesture imitation, including arms motion based on joints‟ movements.

This paper is focused on the programming by imitation, transferring of the motion mapping from

virtual space in 3-D dimensional real physical space. We believe that a deep evaluation of a

robot programming platform is extremely important for the robotic research community since the

technological requirements and the costs to develop more sophisticated programming platform

Page 83: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

83

still prevent them to become generally available. Currently, most high-end programming

platform are developed as prototypes platforms under the supervision of academic community.

Therefore, our platform RMIP such as low-cost programming platform provides an exciting and

inexpensive opportunity for research in programming by imitation.

The programming method, based on the imitation of virtual prototypes, must be still enhanced to

respond at the question: how dynamical virtual system models can be up-dated online to take

dynamic events, from the real environment, into account? The author expect fully automated

robot programming by imitation, using robust enough system to be applied in practical

applications, will not become true before the end of this decade.

ACKNOWLEDGMENT

The author wish to thank, for cooperation and engagement in research activity, the entire team of

Services and Products for Intelligent Environment Laboratory, within the Research &

Development Institute ICDT-PRO-DD of the Transylvania University of Brasov. They granted

us the opportunities to work in team with them and provided us important experiences to extend

this interesting research.

Page 84: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

84

Chapter 4 Modelling and robots behavioural simulation in visual environments

4.1 System concept development with virtual prototyping

Introduction

Through careful development of details, 3D graphics applications can generate virtual prototypes

that provide a useful tool supporting engineering and specialty disciplines analysis for a wide

variety of system developments, both early in the concept development phase and later in

advanced engineering. Virtual prototyping permits users, designers, and logisticians to visualize

two and three dimensional relationships and clearances for joint multi-disciplinary analysis.

This early analysis and identification of design issues offers potential to reduce development

costs since mistakes will be made in the computer rather than on full-scale prototypes. Due to

design maturity, virtual prototypes developed in advanced engineering development will be

based on design analysis and known qualities.

In contrast, exceptional care must be taken during the technology assessment and advanced

concept phases, when details are less certain, to conduct thorough research that produces and

supports a credible product based on expert opinion, technology analysis, and traceable

assumptions rather than “fantasy.” The development of worksheets containing source and

rationale information on each virtual component is a disciplined technique that helps to ensure

reality is not misrepresented.

In manufacturing, rapid prototyping is used to create a three-dimensional model of a part or

product.

In addition to providing 3-D visualization for digitally rendered items, rapid prototyping can be

used to test the efficiency of a part or product design before it is manufactured in larger

quantities. Testing may have more to do with the shape or size of a design, rather than its

strength or durability, because the prototype may not be made of the same material as the final

product. Today, prototypes are often created with additive layer manufacturing technology, also

known as 3-D printing. Direct metal laser sintering (DMLS) may also be used to create

aluminum, stainless steel or titanium prototypes. This process uses laser beams to melt and fuse

metal powders into solid parts.

In software development, a prototype is a rudimentary working model of a product or

information system, usually built for demonstration purposes or as part of the development

process. In the systems development life cycle (SDLC) Prototyping Model, a basic version of the

system is built, tested, and then reworked as necessary until an acceptable prototype is finally

achieved from which the complete system or product can now be developed.

In prototype-based programming, a prototype is an original object; new objects are created by

copying the prototype.

Virtual prototyping, often known as VP, is a software-based engineering discipline which

involves modelling a system, simulating and visualising its behaviour under real-world operating

conditions, and refining its design through an iterative process. The full-motion behaviour of

complex mechanical systems can be analysed before building an actual hardware prototype.

Users can quickly explore multiple design variations, testing and refining until system

Page 85: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

85

performance is optimised. This can help reduce the time and cost of new product development,

whilst significantly improving the quality of overall system designs.

In hardware design, a prototype is a "hand-built" model that represents a manufactured (easily

replicable) product sufficiently for designers to visualize and test the design.

Prototyping Model

The Prototyping Model is a systems development method (SDM) in which a prototype (an early

approximation of a final system or product) is built, tested, and then reworked as necessary until

an acceptable prototype is finally achieved from which the complete system or product can now

be developed. This model works best in scenarios where not all of the project requirements are

known in detail ahead of time. It is an iterative, trial-and-error process that takes place between

the developers and the users.

There are several steps in the Prototyping Model:

1.The new system requirements are defined in as much detail as possible. This usually

involves interviewing a number of users representing all the departments or aspects of the

existing system.

2.A preliminary design is created for the new system.

3.A first prototype of the new system is constructed from the preliminary design. This is

usually a scaled-down system, and represents an approximation of the characteristics of the final

product.

4.The users thoroughly evaluate the first prototype, noting its strengths and weaknesses,

what needs to be added, and what should to be removed. The developer collects and analyzes the

remarks from the users.

5.The first prototype is modified, based on the comments supplied by the users, and a

second prototype of the new system is constructed.

6.The second prototype is evaluated in the same manner as was the first prototype.

7.The preceding steps are iterated as many times as necessary, until the users are satisfied

that the prototype represents the final product desired.

8.The final system is constructed, based on the final prototype.

9.The final system is thoroughly evaluated and tested. Routine maintenance is carried out

on a continuing basis to prevent large-scale failures and to minimize downtime.

The RAD method has a task list and a work breakdown structure that is designed for speed.

However the major difference in RAD is a set of management techniques that are optimized for

speed.

Prototyping - an approach based on creating a demonstrable result as early as possible and

refining that result. The refinement is based on feedback from the business, the eventual users of

the system. Prototyping requires an open approach to development; it also requires an emphasis

on relationship management and change management. There are dangers involved in starting

prototype development too early and in starting it too late.

Rapid application development (RAD) is a concept that products can be developed faster and of

higher quality through:

Gathering requirements using workshops or focus groups

Prototyping and early, reiterative user testing of designs

The re-use of software components

A rigidly paced schedule that defers design improvements to the next product version

Less formality in reviews and other team communication

Page 86: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

86

Object-oriented programming (OOP) is a programming language model organized around

objects rather than "actions" and data rather than logic. Historically, a program has been viewed

as a logical procedure that takes input data, processes it, and produces output data.

Simula was the first object-oriented programming language. Java, Python, C++, Visual

Basic.NET and Ruby are the most popular OOP languages today. The Java programming

language is designed especially for use in distributed applications on corporate networks and the

Internet. Ruby is used in many Web applications. Curl, Smalltalk, Delphi and Eiffel are also

examples of object-oriented programming languages.

Programmers and analysts can build and show visual representations of the designs and

workflow to users.

Prototype that contains only some of the essential system features.

Users can interact with the system.

Prototypes are useful in seeking user reactions, suggestions, innovations, and revision

plans.

Prototyping may be used as an alternative to the systems development life cycle.

Users respond to actual working prototypes.

Analysts refine designed modules based on user responses.

A system is tested and modified as necessary, and later is implemented

In object-oriented programming (OOP), objects are the things you imagine about first in

designing a program and they are also the units of code that are eventually derived from the

process. In between, each object is made into a generic class of object and even more generic

classes are defined so that objects can share models and reuse the class definitions in their code.

Each object is an instance of a particular class or subclass with the class's own methods or

procedures and data variables. An object is what actually runs in the computer.

Visual prototyping

Virtual prototyping with 3D drawing programs provides a means of rapidly developing system

concepts and analyzing them for form, fit logistics, human factors integration, and general

feasibility analysis. The resulting models can be studied, viewed from different angles, and even

“entered” by multidisciplinary design teams working in an integrated product team environment.

Virtual prototyping is an aspect of information technology that permits analysts to examine,

manipulate, and test the form, fit, motion, logistics, and human factors of conceptual designs on

a computer monitor. It facilitates communication between different engineering disciplines

during the early design process, and also provides quality illustrations that help "sell" the design

or program.

A virtual prototype is defined as "A computer-based simulation of a system or subsystem with a

degree of functional realism comparable to a physical prototype" and virtual prototyping as "The

process of using a virtual prototype, in lieu of a physical prototype, for test and evaluation of

specific characteristics of a candidate design.

Virtual prototyping can replace the expensive physical model constructed to test designs. The

value of virtual prototyping is rapidly being recognized for a wide range of engineering

applications. These applications range from illustrating the potential of a system early in the

technology assessment activity or early developmental phase to detailed analysis of mature

designs in the advanced engineering phase. Some of the uses include design form, fit, and

function, logistics and human factors analysis.

Page 87: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

87

Virtual prototypes are developed from concept drawings or blueprints, and contain as much

detail as required to support the immediate need. The same model can be expanded later as more

detail is required. In the computer, the various parts are developed to scale and then are

assembled the same way they would be on a fabrication shop floor. This allows the analyst to

test assembly procedures and human factors as well as the form and fit of the final product. It

also allows parts to be removed to show otherwise hidden components.

The model appears realistic and may be feasible, but it is not backed up with engineering

projections for load bearing features. Taking such liberties during the projection of new

technology into concept systems would be unethical and just poor engineering analysis. To

maintain credibility, virtual prototypes at all stages of engineering development must be based

on disciplined analysis and sound technology projections. Construction details must be backed

up with engineering analysis to provide value and credibility.

Virtual prototypes can enhance the communication of the answers to these questions and provide

a credible baseline for discussion and analysis. In order to support the program, virtual

prototypes must be based on sound engineering and projections of technology growth that are

traceable to sources such as programs developing or producing related hardware, expert opinion,

field and laboratory tests, and engineering analysis. To produce a high quality accurate model of

a technology projection that is credible and useful, each depicted feature of the model must be

traceable to one or more of these sources.

Through careful development of details, 3D graphics applications can generate virtual prototypes

that provide a useful tool supporting engineering and specialty disciplines analysis for a wide

variety of system developments, both early in the concept development phase and later in

advanced engineering. Virtual prototyping permits users, designers, and logisticians to visualize

two and three dimensional relationships and clearances for joint multi-disciplinary analysis. This

early analysis and identification of design issues offers potential to reduce development costs

since mistakes will be made in the computer rather than on full-scale prototypes.

Due to design maturity, virtual prototypes developed in advanced engineering development will

be based on design analysis and known qualities. In contrast, exceptional care must be taken

during the technology assessment and advanced concept phases, when details are less certain, to

conduct thorough research that produces and supports a credible product based on expert

opinion, technology analysis, and traceable assumptions rather than “fantasy.” The development

of worksheets containing source and rationale information on each virtual component is a

disciplined technique that helps to ensure reality is not misrepresented.

Virtual prototyping is a software-based engineering discipline that entails modelling a robotic

system, simulating and visualising its 3D-motion behaviour under real-world operating

conditions, and refining/optimising the design through iterative design studies prior to building

the first physical prototype.

Thus, at its most basic level VP is a tool for enabling engineers, designers and product

developers to work together concurrently within a virtual environment to solve design,

manufacturing and maintainability issues at the earliest stage of product development.

It represents a design capability, which allows users to predict and prevent problems early in the

product-development process rather than finding and fixing them later on, a situation that can

substantially reduce product-development costs.

Page 88: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

88

The adoption of tools that help engineers eliminate product flaws at the earliest stages of

development also helps organisations to meet critical time-to-market objectives, enabling them

to maximise their profit margins.

Virtual prototyping allows engineers and designers to utilise CAD data and techniques to

construct interactive simulations that model the key aspects of the product‟s physical behaviour,

all at the „digital‟ development stage. This allows for product testing at the earliest moment

possible, which has beneficial consequences of the cost of getting the design to market.

Visual prototyping as a technology has evolved in line with technical advances in computing to

become an invaluable tool also in robotic systems area. With the Internet now a powerful global

communications tool, new product visualisation and dynamic testing can be performed online.

Ideally, a virtual robot can be viewed, listened, smelled, and touched by an engineer or a

customer.

This is the area that virtual reality techniques can play an important role. More importantly,

various perspectives of the designed product should be able to tested and evaluated. Visual

prototyping provides a means to quantitatively describe product behavior from various aspects

and thus could possibly be used as a fundamental tool to support a quantitative concurrent

design.

Simulation has been recognized as an important tool in robotics in designing new robots,

investigating their performances and in designing applications of these robots. Simulation allows

us to study the structure‟s characteristics and the functions of a robot systems at different levels

of details, each posturing different requirements for the simulation tools. For example a fast

process, as the moving robot joint, can be slowed down to observe all details in “slow motion”.

All these make things easier and cheaper.

As the complexity of the system under investigation increases, the role of the simulation

becomes more and more important. That's why the simulation tools can certainly enhance the

design development. Depending on the robot application, different structural attributes and

functional parameters have to be modeled

A large amount of simulation software is available for robot systems, and it is already being used

extensively. The majority of the robot simulation tools focus on the motion of the robotic

manipulator in different environments. As the motion simulation has a central role in all

simulation systems; they all include kinematics or dynamic models of robot manipulators. Which

type of models will be used depends on the objective of the simulation technique.

For example, trajectory planning algorithms depend on kinematics models. Similarly, the

construction of a robotized cell can be simulated efficiently by using only kinematics models of

robot manipulators, without considering the dynamics or drives.

On the other hand, to design the actuators, dynamic models are needed. Modern control systems

of robotic manipulators use internally different robot kinematics and dynamic model, to improve

the performances.To model and simulate a robot system different approaches are possible. They

can differ in the way the user builds the model.

Block diagram-oriented simulation software requires that the user describes the system by

combining the blocks but there are other techniques requiring the manual coding.

Page 89: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

89

The simulation tools for robotic systems can be divided into two major groups: tools based on

general simulation systems and special tools. Tools based on general simulation systems are

usually modules libraries or user interfaces, which simplify the building of virtual robot systems

and environments (Fernandez-Madrigal et al., 2007). One of the advantages of such integrated

toolboxes is that they enable us to use other available simulation tools to perform different tasks

such as to design control system, to analyze simulation results, to visualize results, etc.

There exist several general simulation tools which are used for simulation of robot systems like

MATLAB Robotics Toolbox (Corke, 1996), Dymola / Modelica, 20-sim, etc.

Special simulation tools for robots cover one or more fields in robotics like mechanical design,

kinematics (geometry) and dynamics analysis, design of robot work cells, off-line programming.

They can be specialized for special types of robots like mobile robots, underwater robots,

parallel mechanisms, or they are assigned to predefined robot family such as KUKA sim, Fanuc

Simulator etc.

Multi-body dynamic engines

A physics engine is a computer program that simulates the physics models which respect the

laws of physics. It can simulate and predict effects under different conditions that would

approximate what happens in real world.

In robotics, physics engines have two core components, a collision detection system, and the

dynamics simulation system responsible for animating the objects. In the last years new

simulation tools have been available based on general engines for the simulation of physics

environments. These engines provide libraries for simulating the multi-body of an assembly with

constrained or restrained bodies. As examples for libraries to simulate the multi-body dynamics

we have selected the Open Dynamics Engine (ODE) and Open Graphics Library (OpenGL).

ODE is an open source, high performance library for simulating articulated rigid body dynamics.

It is fully featured, stable, mature and platform independent with an easy to use C/C++

Application Programming Interface (API).

ODE is useful for simulating objects in virtual reality environments and virtual robots. With

ODE building the model of a robot is simple.

OpenGL is a standard specification defining a cross-language cross-platform API for writing

applications that produce 2D and 3D computer graphics.

As example to simulate the multi-body dynamics with learning environment we have selected

the BehaivourSim. BehaivourSim is a learning environment for behaviour-based agent. It

supports learning of behaviour-based control by defining simulated agents (Wooldridge et al.,

2002) in an intuitive manner corresponding to the behaviour-based paradigm. Behaviour-based

control is one of the fundamental control paradigms for autonomous agents to realize adaptive

behaviour in a dynamical environment.

However, the diversity of the tasks to be done carries out the study of other types of robots which

may have new properties, whose parameters differ from what is usually known. In order to

enlarge the possibilities of such systems, it may be interesting to relax some constraints on the

parameters using the physics engine for simulation technique.

Modelling and robots behavioural simulation in visual environments

Page 90: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

90

There are two large structural classes of robots: serial robotic structures and parallel robotic

structures. Simplicity considerations in manufacturing and control have led to robots with

revolute or prismatic joints only. Also, on the same considerations joint axes are placed

orthogonal or parallel instead of being placed arbitrarily. Therefore, a variety of simulation tools

has been developed for the robotic systems. The simulation tools are used in mechanical design

of robotic manipulators, design of robot control systems, off-line programming systems etc.

To overcome the problems which arise when the system is very complex - and the robots usually

are very complex systems - several approaches exist to automatically generate the kinematics

and/or dynamic models of robots.

Delphi simulators

In this paper the authors propose the Delphi informatics environment, for the qualitative

simulation of the robotic systems, using the visual programming. The prototype graphical

simulation system has been developed to support and demonstrate different aspects of robots

behaviour.

Delphi is a visual programming tool which makes possible to carry out, in a practically

instantaneous way, the interface with Windows applications. Having a very intuitive access, this

extremely rich tool offers with effortlessness the management of an environment as

sophisticated as Windows.

Delphi language comes from the Pascal, witch is object oriented, just like C++. This one is very

simple to learn, very evolutionary and very intuitive. The simulator created in the Delphi

environment was used to test the performances of the mechanical system and the integrated

environment. The simulator is created with the DELPHI code, which allows us to create the

virtual prototype.

In robotics, the education of engineers demand a well structured understanding of what exactly a

robot is. The virtual robot model gives exactly this understanding. Model driven simulation is a

valuable tool for understanding the properties of a robot. Delphi system encourages users to

make their own program expansions and simulate them. Doing this, new ideas can be tested on

the robots or on the production systems.

In general, the simulation can be used to perform analysis and design studies on any robot

system which can be modelled as a set of rigid bodies interconnected by joints, influenced by

forces, driven by prescribed motions, and restricted by constraints.

For simulation in Delphi, one supposes the pure motion, without reference to the masses or

forces involved in it. It is the case of the kinematics model that studies the motion of a robot arm,

without considering its mass or the forces which are acting on it. This model deals with the

characteristics of motion, without regarding the effects of forces or mass (Fleury et al., 2006).

In traditional robotics, programmers attempted to anticipate and explicitly control every aspect of

the action of the robot. However, these systems often failed when are placed in a changing

environment, where unanticipated situations appear. Systems using autonomous agents, with de-

centralized control of robot motion, enable robots to learn and adapt to changing environments.

Autonomous agents are software programs designed to pursue goals and can do so without user

intervention. They are usually large applications, which attempt to solve practical industrial

Page 91: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

91

problems, and require a large amount of investment and development. Advanced robotic systems

involve large agent-based simulators aimed at understanding "action selection". By this

mechanism the robot selects the behaviour to execute from much possible behaviour.

Delphi simulators are generally smaller programs, and they are written in order to extend

research in Robotics. They usually involve algorithms that allow exposing the behaviour of the

virtual robots and their environment. This technique can also be used to provide a method of

robotic system command.

All of these programs enable the manifestation of global behaviours. They also attempt to

simulate more aspects of robot behaviour - not just movement. As such they include the

movement of robot manipulators, the contact with manipulated objects and the collision with

unexpected objects into the work environment. The information regarding the collision are given

about the shape of each body and then it figures out which bodies touch each other.

The Delphi programs are written with an object-oriented approach. The simulation is driven by a

special object, the Object Inspector. The Object Inspector maintains a list of the active object

and sends step messages when it is time for objects to update themselves. All components of the

artificial world are objects, as is the environment itself. The interface consists of a collection of

tools for analysis and visualization.

Data analysis objects can communicate with the user interface objects to present displays of data,

as well as for storage to files. The prototype allows the user to create, modify, and destroy

objects. Each object has a few standard attributes managed by Object Inspector, as well as user-

specified private data.

The requirements of the simulation to support such research activity are:

This simulation should be able to show visually how behaviour can evolve.

It will allow the testing of different theories of robots control, and should enable

experimentation in the “hardware-in-the-loop” simulation mode.

The aim of Delphi-simulation is to contribute to a better understanding of robot behaviour.

Delphi-simulation can be used to produce virtual applications, making them accessible to new

kinds of experimental manipulation and testing. The simulator should be able to present a

changing environment in which robotic systems are continually changing over time.

The Delphi system uses a library tool, and a system to run the simulation. The simulators are

still under development, although prototype versions have been produced. We present some

typical simulation examples in different fields of robotics.

Page 92: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

92

4.2 Virtual Prototyping for robotic systems

The purpose of the paper is to emphasize the role of the simulation in different fields of robotics.

Also we recommend to researchers to use the general dynamic engines and general visual

programming languages for the simulation and visualization of robotic systems.

This paper describes the development of a robotic language and the creation of a virtual reality

system for the simulation of robotic systems. We propose a description language for specifying

motions and for allowing robots to acquire motor skills. Locomotion greatly increases our ability

to interact with our environments, which in turn increases our mental abilities. This principle

also applies to intelligent robots. However, there are great difficulties to specify path motion and

to represent motor skills, which in most cases require four-dimensional space representations.

We propose a representation framework that includes the following attributes: motion

description layers, reference system, progressive quantized refinement, and automatic constraint

satisfaction. We also outline strategies for acquiring new motor skills by learning from trial and

error, virtual motion path, and programming. This paper discusses the using of the visual

programming Delphi environment, for facilitating intuitive robot programming. From a robotic

point of view, the virtual robotic structures created were simplified and their analyzing allows a

better description of the behaviour of real robotic structures.

The simulations prepared for this application were also made for the development of the

systematic methods for creating the new robotic systems.

The results of this investigation consist in an identification of the properties and constraints of

the virtual robot prototypes. The virtual structures created in this paper encompass the behaviour

of robotic systems. They may present a behaviour which is qualitatively different from the

behaviour that the robots currently have.

In the end, the simulation in robotics has reached a very important role and by using different

simulation software, the present and future capabilities of complex robotic systems can be

significantly improved.

The simulation is now a powerful tool supporting the design, planning, analysis, and decisions in

different areas of research and development. Simulation has become a strategic tool in robotics

as a modern technological branch. The simulation is very common in our lives to understand

reality in all its complexity. We try to build artificial objects/systems and to animate them to

understand their behaviour.

Advanced simulation tools are the foundation for the design of sophisticated robot systems, for

the application of robots in complex environments and for the development of new control

strategies and algorithms.

The simulation, being once a tool for the analysis of the robotic systems and the task planning,

has become a technique for developing new robot systems. Not only that the modern simulation

Page 93: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

93

tools can simulate and visualize the real world in a very realistic way, but also they allow us to

see beyond the reality.

Advanced robot systems require sophisticated simulation tools which can simulate exactly

enough the physical world, at sufficient speed and allow users interaction. New challenges in the

simulation of robotic systems are multi-body dynamic simulation techniques that compute the

robot and the manipulated object motion under the influence of external forces, collision

detection, contact determination and realistic visualization of the robot and its environment

interaction.

Simulation is the process of a model‟s design of a real or theoretical system. The execution of a

model and its analyzing is necessary before the system‟s implementation. Actually, in robotics

simulation plays a very important role, possibly more important than in many other fields and we

like to present in the following some approaching in the robotics from the behavioural simulation

point of view.

The principle “learning” is essential in simulation. Using simulation we can learn about some

things in a very effective way and while modifying “rules” we can observe the effects of our

interaction. As it is in our nature that “seeing is believing”, the visualization is the other

dominant in simulation.

Simulation is a highly interdisciplinary field since it is widely used, in all fields of research -

including the robotics - and at different levels, from academic research to manufactures. Being

able to simulate, opens a wide range of options for solving many problems creatively. We can

investigate, design, visualize and test an object even if it does not exist. We can see the

performances of a system before it is built. It is possible that our solutions may fail or even blow

up, but only in simulating on the virtual prototype (Bruzzone et al., 2003).

So, using the simulation tools one can avoid injuries and damages, unnecessary changes in

design, after the production has already started a long cycle manufacturing process. Using the

virtual prototypes one can avoid even unnecessary paper works. With up-to-date simulation tools

one can deal with exact geometry, consider the dynamic characteristics of a system, include the

man–machine interfaces, and visualize objects in 3D in detail.

Having all these in mind, for what is possible or not, the boundaries are pushed far away

especially with advanced virtual reality tools. Using simulators, the researchers may build

experimental environments, according to their own imagination. What they have imagined

during the night, it can be transposed into visual imagines the next day. Complexity, specificity

can be gradually increased to a level where virtual systems can head to real challenges of the

physical world.

Multi-body serial robot systems

In its most general form, a serial robot system consists of a number of rigid links connected with

joints. The position and orientation of a robot's end-effector are derived from the joint positions

of the robot arm.

For first example, we use a virtual serial robot, which manipulate a container. We have to create

all bodies and connect them if desired with proper joints. For example, the 3DOF model as

shown below in Fig.4.2.1, integrated with a conveyors system. Here, the multi-body library

provides three-dimensional mechanical components to model rigid multi-body robot system.

Page 94: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

94

The robot system is built by connecting blocks representing parts of the robot like link bodies,

joints, actuators and gripper.

The simulation methodology uses a number of possible scenarios and involves some sequences

of animation, as one can see in the Fig.4.2.1 and Fig. 4.2.2. Fig.4.2.1 shows how a robot hand

can grasp a container which is moving. In this example, the synchronization between robot arm

and conveyor is studied and all contacts between the fingers of the gripper and the container are

analyzed.

Fig.4.2.1 Container grasping scenarios

The dynamic simulation of multi-body systems becomes very important when the robot

manipulator must interact with the mobile objects, where the success will depend only on the

capabilities of the robots. In our example, the collaborative work between transport system and

robot hand is analyzed. This allows us to simulate an entire grasping task, as well as test robot

control algorithms.

For the distribution task in the Fig. 4.2.2 is presented the three posture of the scenarios created to

depose the container on the corresponding conveyor. The criterion, after which the conveyor is

selected, is the colour of both container and conveyor. The three system configurations are

presented below for three different colours: blue, green and maroon.

Fig.4.2.2 Scenario for the container placement on the corresponding conveyor according to

container color

Multi-body parallel robot systems

Parallel mechanisms generally comprise two platforms that are connected by joints or legs acting

in parallel. For the parallel robot applications, as an example we have selected the hexapod

platform.

Page 95: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

95

The hexapods are complex mechanisms made up of six (often identical) kinematics chains, of a

base (fixed rigid body including six joints or articulations) and of a platform (mobile rigid body

containing six other joints).

.

Fig.4.2.3. Triangular moving platform and double hexapod platform

The proposed model with double platform contributes at the increase of the stability of the upper

platform by compensating the perturbation with double group of active legs.

Implication of the behavioural simulation in the control design

A very important part of the robotic system is the control system. From the control viewpoint

there are different control levels. The lowest level is the close-loop control and the higher is the

trajectory planning. The path planning and other more global control tasks are performed at

higher levels.

In the process of controller design different steps have to be performed. First of all, the system

has to be modeled. In the next step, the control algorithm is developed. The first results are then

obtained by the simulation. If the results are satisfactory, then in the final step the control

algorithms are tested on a real system. For this, a real-time code should be generated and

implemented on the real system.

In an integrated environment one can sustain two modes of operation. In the “pure” simulation

mode the virtual robot system is experimented. In the “hardware-in-the-loop” simulation mode,

some of the real system (e.g. a real robot with sensors) is included in the simulation loop. Pre-

requisites for the “hardware-in-the-loop” simulation are the real-time simulation ability and the

necessity of additional hardware components, which are needed for linking together the real

system with her simulator. The integration of the two modes of operation is the most important

feature of the integrated environment.

To taste the robot control algorithms, in this paper we have developed the Delphi simulators for

virtual robot systems. (Fratu and Fratu, 2008). The informatics system was developed specially

for testing the different control algorithms for a real grasping process.

Using the virtual system, the controller can be tested without a real robot, e.g. before even the

robot has been built. When the real robotic system is tested, the connection between real control

unit and the simulator system must be done. If the real robot system imitates the virtual robotic

system, will be attended the ideal control.

Page 96: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

96

Conclusion

The purpose of the paper is to emphasize the role of the simulation in different fields of robotics.

Also we recommend to researchers to use the general dynamic engines and general visual

programming languages for the simulation and visualization of robotic systems.

This paper discusses the using of the visual programming Delphi environment, for facilitating

intuitive robot programming. From a robotic point of view, the virtual robotic structures created

were simplified and their analyzing allows a better description of the behavior of real robotic

structures.

The simulations prepared for this application were also made for the development of the

systematic methods for creating the new robotic systems.

The results of this investigation consist in an identification of the properties and constraints of

the virtual robot prototypes.The virtual structures created in this paper encompass the behaviour

of robotic systems. They may present a behaviour which is qualitatively different from the

behaviour that the robots currently have.

In the end, the simulation in robotics has reached a very important role and by using different

simulation software, the present and future capabilities of complex robotic systems can be

significantly improved.

Another article published in the System concept development with virtual prototyping topic is

Simulation of articulated robots for virtual prototyping in dynamic 3D environments.

This paper deals with the simulation of a dynamical system in which the motion of each rigid

robot is subject to the influence of virtual forces induced by geometric constraints. These

constraints may impose joint connectivity and angle limits for articulated robots, spatial

relationships between multiple collaborative robots, or have a robot follow an estimated path to

perform certain tasks in a cycle. In this paper the authors give a brief overview of a general

simulation framework, describing the primary tasks which a simulator needs to implement. The

robot behavioral simulation in the virtual environment enables us to predict the behavior of a

given real manipulator into real environment. The robot behavioural simulation enables us to

predict the behavior of a given manipulator under given initial conditions, applied torques, and

applied loads.

The ability of predicting this behaviours is important for several reasons: for example, in design

the designers want to know whether with a given selection of actuators, the manipulator will be

able to perform a certain typical task in a given time frame; in creating feedback control

schemes, where stability is a major concern, the control engineer cannot risk a valuable piece of

equipment by exposing it to untested control strategies. Therefore, a facility capable of

predicting the behaviour of a robotic manipulator, or of a system at whole, for that matter,

becomes imperative.

In this paper, the authors present a new motion planning algorithm for virtual prototyping. This

algorithmic structure is inspired by constrained dynamics in physically-based modelling.

Page 97: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

97

The authors seek to deduce a virtual geometry of the objects, a 3D geometric realization of a

collection of rigid bodies is visible in the drawing. The authors transform the motion planning

problem into a dynamical system simulation by treating each robot as a rigid body or a collection

of rigid bodies moving under the influence of all types of constraint forces in the virtual

prototyping environment.

These may include constraints to enforce joint connectivity and angle limits for articulated

robots, constraints to enforce a spatial relationship between multiple collaborative robots,

constraints to avoid obstacles and self-collision, or constraints to have the robot follow an

estimated path to perform certain tasks in a cycle.

Proposed constraint-based planning structure has the following characteristics:

It can handle both static environments with complete geometric information or dynamic

scenes with moving obstacles whose motion is not known a priori.

It is applicable to both rigid and articulated robots of arbitrarily high degrees of freedom,

as well as multiple collaborative agents.

It allows specification of various types of geometric constraints.

It runs in real time for modestly complex environments.

The authors demonstrate the effectiveness of this structure for the problem of virtual assembly

prototyping with applications in assembly line planning,

Robots’ Motion Simulation

The robots‟ motion should be animated with the highest degree of realism possible using motion

capture data or accurate full-body simulation, while the multitudes secondary details to the

auxiliary elements (scene, cameras etc. ) can be simulated at much lower fidelity.

The classic robot motion problem, also referred to as the Piano Mover‟s problem, can be stated

as the following: given a robot R and a workspace W, find a path from an initial configuration I

to a goal configuration G, such that R never collides with any obstacle Oi from a set of obstacles

O along the path P, if such a path exists.

The path P is a continuous sequence of positions and orientations of R. Continuous sequences of

positions and orientations of R are assimilated with the robot system animation on a virtual

scene. Despite the exciting progress in the field, simulating a dynamical system with many

degrees of freedom remains a computational challenge. One of the central components of any

control or simulation system for articulated bodies is forward dynamics [3].

Forward dynamics computes the acceleration and the resulting motion of each link, based on the

given set of external forces and active joint forces. The known algorithms have a linear- time

dependence of the number of degrees of freedom. This permits any object in a scene to behave in

a physically-plausible way: they accelerate, recognize collisions, and respond to collisions much

like one would expect it to respond.

Several techniques have been proposed for accelerating various types of dynamic simulation.

Yet, there exists no known general algorithm for automatic simulation of articulated body

dynamics.

Plausible motion simulation

Reasearcher Barzel introduced the idea of “plausible” motion, i.e. motion that could happen and

look physically plausible to the viewers. For many visual applications or real-time interaction,

accurately simulating all the details of the real environment is not necessary.

Page 98: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

98

In fact, it is often sufficient to provide effective motion to make the scene appear more realistic,

without committing much computational resources.

In an environment with uncertainty, we generally expect a constrained problem to have multiple

solutions. It is difficult to know before what solutions are available. Hence, it is bad to use a

solution strategy that seeks a single answer; rather, it prefers a technique that produces many

solutions that reflect the range of possible outcomes. While for feature animation a user is

expected to choose the one animation they prefer, other applications benefit directly from

multiple solutions:

Computer simulator designers can use different animations each time a simulation is on

stage, making it less predictable and potentially more entertaining.

Training environments can present trainees with multiple physically consistent scenarios

that reflect the physics and variety of the real world.

The authors generate multiple animations that satisfy constraints by applying an original

algorithm to trial from a randomized model [6]. The algorithm needs the model of the

environment, including the sources of uncertainty and the simulator that will generate an

animation in the virtual environment. The algorithm described in this paper generates an

arbitrarily sequence of animations in which “good” animations are expected to appear.

Simulation loops

The simulation loops fix robots‟ components algorithmically, in order to generate physically

plausible motion. Note that not all the pieces are put together into a unified system.

Many related works describes a simulation core or a simulation loop to achieve this [5]. The

following are the general steps in a simulation core:

1. Clear the force accumulators: Each body or its components maintains a total of all forces on

it. At the beginning of each step, we clear the forces from the previous step.

2. Detect collisions: Loop over all bodies and determine any contacts or collisions in the scene

and prepare them to be resolved.

3. Compute external forces: Loop over all external forces, including contact or collision

resolution forces, and add them to the force accumulators.

4. Compute constraint forces: At this point, each body or component knows the total force acting

upon it (it‟s in force accumulator). To process the constraints, first add any soft constraints to the

accumulators. Finally, compute and apply the hard constraint forces.

5. Compute derivatives: Gather derivatives as needed to prepare to update the system.

6. Integrate and update the state of the simulation: Use a numerical integration method to update

the state of the system by some small time increment.

Implementation

Our system was implemented with DELPHI object-oriented programming language. The authors

used in-house library ANIMATION-VIEW for collision detection by generating of the distance

fields for surface repulsion constraints.

Platform‟ toolbox offers the Delphi functions for the implementation of the virtual system

prototypes. For discrete set {tk} of instants, Delphi system generates an images sequence of the

virtual robot system.

System Demonstration

Page 99: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

99

The authors have tested the proposed motion planning system in the following virtual

prototyping application: Assembly Line Planning. An animation generated from this type of

scenario is shown in Figure 4.

Fig.4.2.4 Assembly Line Planning Scene

The robot arms avoid the moving object to reach a moving part passing on the conveyer belt. In

this scenario the aim is to animate all actors to realize the assembly of the automobile. In this

example, shown in Figure4, the robot arms from scene must access a part moving and past it on a

conveyer belt.

The factory floor contains a transfer structure that is moving over the conveyer belt in the

opposite direction to the part‟s movement. The moving obstruction causes the robots to

reactively modify its path to avoid collision.

In this example, constraints can be defined for any aspect of the object‟s 3D state at any point in

time. Initial conditions for the object are specified by constraining its state at the start of the

simulation.

Conclusion

The authors have presented a novel framework for motion planning in virtual prototyping

applications. Thy have reformulate the motion planning problem into a virtual simulation

problem where constraints on the robot‟s motion guide it from its starting configuration to its

target, on the virtual scene. These constraints can‟t impose penetration constraints among

objects, the angle limits and connectivity of articulated robot joints. The avoidance of collision,

the following of estimated paths, and many other possible relationships between the cooperative

robots and objects on the scene, are feasible in the virtual environment.

The models proposed by authors arise naturally in the virtual environment and provide a means

of verifying the plausibility of the motion in the real environment. With further work it should be

possible to experimentally obtain more accurate robot dynamically models who require finding

good animation.

Page 100: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

100

Acknowledgments

This work was supported in part by LAMIH - Laboratoire d'Automatique, de Mécanique, et

d'Informatique industrielles et Humaines - UMR CNRS 8530 - of the Valenciennes University -

France. The authors thankfully reply for the support of this institution.

(B-ii) The evolution and development plans for career development

Research Directions

Based on my past experience and on the research and teaching directions detailed below, my

short term goal is to build a research team that can accumulate expertise in most areas related to

haptic interfaces and robot collision avoidance. All my research and teaching activities are

closely intertwined, and this could be a catalyst for the team to become competitive in applying

for national and European funding.

The long term goal is to lead a group that could produce high quality research - published in

competitive venues and highly cited, but could also produce prototypes and experimental

systems to bridge collaboration with the industry. I am presently contributing to one projects

collaborating closely with foreign academic partners (University Paris- Sud France), and in

parallel, I am applying for funding in partnership with young research teams - PN-II-RU-TE-

2014-4.

The planning of future research relies on the direction namely Interference between virtual and

physical world as result of our precedent research domains. Various domains can benefit from

the outcome of the research, e.g. robotic systems, learning technologies, medicine and games. I

will refer in the following to two future research topics I would also like to propose to the

potential Ph.D. students. They also constitute draft applications for national/European funding

grants.

Innovative control methods for haptic interfaces

Principally, Haptic interfaces are based on this interference between these two worlds, virtual

and physical world. Cooperation between virtual and physical world is a field of Artificial

Intelligence concerned with understanding, recognizing, and utilizing human tactile feel in the

design of haptic systems. It has become a very hot research topic in human interaction with

virtual environment via computer, because it helps improve the quality and effectiveness of

human to robots clone communications.

Despite the advances in computing and related technologies, there is no artificial system which

can match the human capabilities in dealing with tactile feel. The aim of the future project is to

investigate and promote new bio-inspired solutions, in the implementation of artificial systems

that can process tactile feel.

The field of Haptic interfaces is planned to be developed, within the Department of Automatics

and Technology of Information of Transilvania of Brasov, an experimental platform within the

further cooperation with our partners from LAMIH Laboratorz from Valenciennes, France and

finding new ones and by involving students in research.

On this platform we propose a description language for specifying motions for haptic robots and

for allowing robots to acquire tactile sensation / feeling.

Page 101: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

101

For haptic application, tactile sensation / feeling greatly increases the ability to interact with

physical environments, which in turn increases mental abilities. This principle is applied to

haptic robots which extend this ability also for virtual environments. However, there are great

difficulties to specify haptic feeling in contact with virtual objects and to represent the motion of

the virtual devices, which in most cases require 3-D dimensional representations.

Currently most humanoid simulation systems require manual manipulations of body parts or

require capturing movements performed by a person. We aim to describe haptic robot motions

using our high level motion description language. We will propose new motion primitives and

new syntax that allows programmers to easily describe complex robot motions. To make the

motion specification possible, we will define a framework that models haptic robots by

specifying their capacities and limitations.

Furthermore, we will develop a simulation system that can execute the motion description

language and can automatically handle conflicting motion statement and can ensure that the

described motions are within the limitations of haptic robots. Our simulation results obtained till

now show that the proposed system has great future potentials.

While computer user interface progressed from text only input to graphical user interface; the

creation of the Virtual Reality(VR) technology allows the interaction between virtual and

physical world to become more intuitive.

On-line collision avoidance using Programming by Demonstration

The main objective of this research is to propose and investigate new methods aiming robot

collision avoidance. We propose a new strategy to robot collision avoidance using programming

imitation paradigm. The concept is based on original idea. His suggestion was that the physical

robot can predict the situation (position and orientation) using its virtual prototype in parallel

with than physical sensory signals in redundancy manner.

For a system with a demand of reacting as precisely as possible, its past information is not

suitable for control planning any more. We should predict the future collision, at the time when

the control command arrives at the physical robot and is executed.

The ability of predicting of the robots collision is important in design; the designers want to

know whether the robot will be able to perform a typical task in a given time frame into a space

with constraints.

The control engineer cannot risk a valuable piece of equipment by exposing it to untested control

strategies. Therefore, a facile strategy for contact detection and collision avoidance, capable of

predicting the behavior of robotic manipulators, becomes imperative.

When the robots need to interact with their surrounding, it is important that the computer can

simulate the interactions of the participants, with the passive or active changing environment in

the graphics field, using virtual prototyping.

For effective involved path navigation, a technique is needed which can exploit the reference

trajectory structure to search in the local continuum for actions which minimize path deviation

and avoid obstacles. A pursuit planner is adequate in order to permit relaxation of a trajectory

(for optimization reasons) by searching a small number of degrees of freedom.

The accepted classes of “pursuit” algorithms will be used to generate a path planner in order to

accomplish robotic tasks, the actions for the each robotic task are computed for virtual robot and

Page 102: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

102

are transferred, with a central coordination to corresponding physical robot which must imitate

its virtual homonym.

In this way, distinct other strategies, our system based on patented solution don‟t require direct

physical contact or exact positioning in front of the obstacle. The virtual robot may interact and

may interpenetrate each with other or with the obstacles and will advise her physical homonyms

before these will crash. Based on save frontier concept, the physical robot will stop in front of

obstacle even if her virtual homonym will penetrate through obstacle.

To program the desired motion sequence for the physical robot, one captures the motion

reference paths from her virtual robot model and maps these to the joint settings of the physical

robot. Motion imitation requires transfer of a dynamical signature of a movement of the virtual

robot to the physical robot, i.e. the robots should be able to encode and reproduce a particular

path. Furthermore, the virtual robot must cover all possible contexts - including la presence of

accidental obstacles - in which the physical robot will need to generate similar motions in unseen

context.

The programming method, based on the imitation of virtual prototypes, must be still enhanced to

respond at the question: how dynamical virtual system models can be up-dated online to take

dynamic events, from the real environment, into account? The author expect fully automated

robot programming by imitation, using robust enough system to be applied in practical

applications, will not become true before the end of this decade.

The advantages of such programming approach as an alternative to the classical methods (e.g.

vision guided trajectory imitation) are on-line adaptation to the motion of the virtual prototype.

Teaching Directions

Successful research activity is not possible without cooperation. Up to now, we have cooperated

in the field of haptic interfaces with a team from the University of Valenciennes and in the field

of robot collision avoidance with a team from the University of Reims, France.

The cooperation with University of Valenciennes I has both a teaching side and a research side

in the framework of visiting professor / researcher mobility.

Within the teaching side I have sustained eight hour-courses for several years with topics from

the robot motion control. This teaching cooperation was in the framework of the Socrates /

Erasmus European program and is continuing and planned to be developed with undergraduate

students and PHD students.

On the other hand, the research cooperation as visiting researcher in the field of haptic interfaces

had as a result the publication of five scientific papers in refereed journals and proceedings of

international conferences in the last ten years.

This very good start is supposed to be continued also in a formal framework. The field of tactile

sense based on intelligent sensors will be approached in the next future.

The presented research directions and topics in robotic, together with existing equipment and

computing facilities constitute a sufficiently large field to accommodate fulltime researchers,

doctoral students and diploma students.

Page 103: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

103

Funding can be ensured through national and European grants and contracts with economic

partners in the geographical area, where all industry now uses intelligent systems as haptic

interfaces systems, manufacturing robotic lines and components and is interested in signal

integrity issues.

A normal enhancement of teaching facilities is expected in the future. However, technical

knowledge acquired at the first two levels in the field of servo-systems for robotic systems is not

enough for performing research activities.

Therefore additional courses must be planned at the doctoral school level. At least two courses

have to be introduced in order to ensure a smooth transition from school activity towards

research: 1. Intelligent motion control and 2. Visual programming.

After finishing the introductory courses, doctoral students are supposed to gain experience and

obtain results by working in the research laboratory under the supervision of and interacting with

the doctoral supervisor and with his/her colleagues in a team-work environment.

Practical experience shows that large research structures are not efficient in particular situation.

We will rather consider formation of small ad-hoc groups or teams created for solving problems

and for application for grants in specific research targets.

From the above presentation, it is clear that the main research topics are planed to belong to the

field of Applied Electrical engineering. However our experience in the other two fields where we

have worked and obtained results, namely robot motion control and modeling and simulation of

the robotic systems will have an important impact on our future development.

For the future, I see the development of a “Interference between virtual and physical world”

curricula as a high priority for the doctoral students. The skill set necessary for this discipline

spans across electrical engineering (servo-motors, electronics, informatics), and by its nature

draws from several areas of technology.

The scientific results achieved during the research will be submitted for publication in ISI

journals and conferences. Parts of the research could be also disseminated in my taught courses

of Simulation of the robotic systems and Robotic Control Systems, recursive positioning

strategy will proposed to be studied at the doctoral school level.

Based on current trends of increasing importance of intelligent robotic systems, the marketplace

is likely to continue to require graduates highly specialized in this area.

Page 104: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

104

(B-iii) Bibliography

1. (1994) Fratu, A.: Cercetari privind sistemul de pozitionare cu servomotor sincron si

microcalculator pentru roboti industriali. Rezumatul tezei de doctorat. Universitatea

Transilvania din Braşov, 1994.

2. (1994) Fratu, A. - Ondulatoare electrice - structuri de comandă pentru roboţi

industriali. Lucrările simpozionului de electronică, Piteşti, 11 - 12 noiembrie, 1994, p.

88 - 91.

3. (1994) Fratu, A. - Metodă de poziţionare a elementelor structurilor de manipulare.

Lucrările simpozionului naţional de Roboţi industriali, vol. 1, 16 - 19 noiembrie, 1994,

Timişoara, p.223 - 226.

4. (1998) Fratu,A. - Forma generală a comenzii unui sistem robot. Lucrările

Simpozionului Naţional de Electronică şi Automatică, Academia Tehnică Militară,

noiembrie, 1998, Braşov.

5. (1998) Fratu, A. - Controlul forţei în structurile de manipulare. Lucrările celui de al

XIV-lea Simpozion Naţional ROBOTICA‟98, 5 - 7 noiembrie, 1998, Braşov, p. 275 -

278.

6. (1999) Fratu,A. – Modelul dinamic si ecuatie de miscare scrisa in spatiul de eroare.

Academia Aviatiei si Apararii Antiaeriene Henri Coanda. Buletinul stiintific, Stiinte

tehnice-Automatica, Anul I, No.10(10) dec. 1999, ISSN nr. 1453-0139, p. 87-90.

7. (2002) Rab Laura, Fratu Mariana s.a.: Perspectives and trends as regards flow

measurements for the automotives sensors, RECENT, Vol. 3 (2002), Nr. 1(6), p. 52-

55.

8. (2002) Fratu, A.,s.a.: Possible Approaches to Problem of Robot Control in Assembly

Operations, RECENT, Vol. 3 (2002), Nr.1(6), p. 33-37.

9. (2003) Fratu, A., s.a.: Generarea miscarii unui brat robot, Revista RECENT, ISSN

1582 – 0246, Vol.4 (2003 ), Nr.1(7), p. 21-25.

10. (2004) Fratu, A., s.a.: Adaptive control for robots, Revista RECENT, ISSN 1582 –

0246, Vol.5( 2004 ). Nr.3(12), p. 34-36.

11. (2005) Fratu, A. et al.: The problem of operational space control for redundant

robots, Revue RECENT, ISSN 1582 – 0246, Vol.6 ( 2005 ). Nr.3 (15), p. 52-55.

12. (2005) Fratu, A., s.a.: Dynamics model of robots, Revista RECENT, ISSN 1582 –

0246, Vol.6( 2005). Nr.2 (14), p. 31-34.

13. (2005) Fratu, A., s.a.: Application‟s development with SCADA system, Revista

RECENT, ISSN 1582 – 0246, Vol.6 ( 2005 ). Nr.2 (14), p. 35 -40.

14. (2005) Fratu, A., s.a.: The problem of operational space control for redundant robots,

Revista RECENT, ISSN 1582 – 0246, Vol.6 ( 2005 ). Nr.3 (15), p. 52-55.

15. (2005) Fratu, A.: Progress and Trends in the University-Industry Co-operation for

foreign Students received at the Automatics Department. Bulletin of the University of

Brasov, ISSN 1223-9631, Vol.12(47)-New Series, 2005, p.557-558.

16. (2005) Fratu, A. et al. : Artificial Inteligence and Inteligences Robots’ apereance,

(Inteligenta Artificiala si aparitia Robotilor Inteligenti), Vol. '' Hermeneutica Historiae

Page 105: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

105

et Philosophiae Technicae '', TRANSILVANIA University Pubishing, ISBN 973-635-

744-9, Liber III, Anno 2005, p,150-154.

17. (2006) Fratu, A et al.: Feedback linearization in robot’s technique, Revista RECENT,

ISSN 1582 – 0246, Vol.7 ( 2006 ). Nr.1 (16), p. 24-27.

18. (2006) Fratu, A. et al.: Repetitive Control Motion in Robots Manipulators’

Technique, Revue RECENT, ISSN 1582 – 0246, Vol.7, Nr.2 (17), p. 37-41.

19. (2007) Fratu,M. and Fratu, A. : Anomaly diagnosis system based on qualitative

models. Transylvania University Publishing, ISSN 1843-6617, CIB-2007, p.551-54.

20. (2007) Fratu,M. and Fratu, A. : Anomaly diagnosis method based on neural networks.

Transylvania University of Brasov Publishing, ISSN 1843-6617, CIB-2007, P.55-58.

21. (2007) Fratu, A., et al.: Avoidance Obstacle Technique Using Redundant Robots,

Revue RECENT, ISSN 1582 – 0246, Vol.8, Nr.1 (19), p. 22-26.

22. (2007) Fratu, A., Vermeiren, L. : Adaptive control architecture for redundant robots,

Revue RECENT, ISSN 1582 – 0246, Vol.8, Nr.2 (20), p. 97-100.

23. (2007) Fratu,A et al.: The science and the art-together-in creating automatically

virtual images. Transylvania University Publishing, ISSN 1843-4959, Vol. ARTA SI

STIINTA, 2007, p.40-47.

24. (2008) Fratu A., Renaux, D., Fratu Mariana: Simulation and Control of the Parallel

Robots‟ Dynamics. Bulletin of the TRANSILVANIA University of Brasov, ISSN

2065-2119, VOL.1 (50), Series I, p.343-350.

25. (2008) Fratu, M. and Fratu, A. : Modelarea defectelor in sistemele de transport

combustibili energetici. Transylvania University Publishing, ISSN 1843-6617, CIB-

2008, p.61-64.

26. (2009) Fratu M. and Fratu, A.: Simulation of the Gas Transportation through Pipes

Inspected by Intelligent Robots. BULETINUL Universităţii Petrol – Gaze din

Ploieşti, Vol. LXI, No. 2, p. 39 – 44, Seria Tehnică.

27. (2009) Fratu A., Brethe. J.F., Fratu M.: Industrial robots in a flexible assembly

system, Revista RECENT, ISSN 1582 – 0246, Vol.10, Nr.1 (25), p. 28-31.

28. (2009) Fratu M. et Fratu, A.: Simulation of the Gas Transportation through Pipes.

Revista RECENT, ISSN 1582 – 0246, Vol.10, Nr.2 (26), p. 105-108.

29. (2010) Fratu, A., Brethe, J.F., Fratu, M.: Redundant inverse kinematics system for

obstacles avoidance, Revue RECENT, ISSN 1582 – 0246, Vol. 11, no. 1(28), pp. 29-

32.

30. (2011) Fratu, A. and Fratu, M. : Analytical Model of the Cutting Process with

Scissors-Robot for Haptic Simulation. Bulletin of the Transilvania University of

Braşov, ISSN 2065 - 2119, Series I: Engineering Sciences, Vol. 4 (53) No. 1 – 2011,

pp. 113-120.

31. (2011) Fratu, A.: Collision Prevention Method for a Dynamic Group of Cooperative

Robots Who Communicate Wirelessly, Revue RECENT, ISSN 1582 – 0246, Vol. 12,

No. 2 (32), pp.131-134.

32. (2011) Fratu, A., Fratu, M.: Collision Prevention Protocol for a Dynamic Group of

Asynchronous Cooperative Robots, Revue RECENT, ISSN 1582 – 0246, Vol. 12 , No.

2 (32), pp.135-138.

33. (2012) Fratu, A., Fratu, M.: Simulation of Articulated Robots for Virtual Prototyping

in Dynamic 3D Environments, Bulletin of the Transilvania University of Brasov,

Series I: Engineering Sciences, ISSN 2065-2119, Vol. 5 (54) No. 1 – 2012, pp.71-76,

indexed in EBSCO

34. (2012) Fratu, A., Fratu, M.: Equation of Motion of the Nonlinear Robotic Systems

with Constraints. Bulletin of the Transilvania University of Brasov, Series I:

Engineering Sciences, ISSN 2065-2119, Vol. 5 (55) No. 2 – 2012, Accepted for

publication, indexed in EBSCO

Page 106: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

106

35. (2012) Fratu, A. and Fratu, M.: Robot Control for Avoid Self-Collisions, Revue

RECENT, ISSN 1582 – 0246, Vol. 13 (2012), No. 1 (34), March 2012, pp.47-50,

Indexed by ULRICHSWEB and Copernicus Journal Master List

36. (2012) Fratu M., Fratu, A.: Collision Avoidance Control Using Null Space Criteria,

Revue RECENT, ISSN 1582 – 0246, Vol. 13 (2012), No. 1 (34), March 2012, pp. 51-

54

37. (2012) Fratu, A., Fratu, M.: Tracking Control of Robotic Multi-Body Systems.

RECENT Journal - ISSN 1582-0246, Vol. 13 (2012), No. 2 (35), July 2012, pp.173-

176

38. (2012) Fratu, A., Fratu, M.: Space Management in Smart Room, RECENT Journal -

ISSN 1582-0246, Vol. 13 (2012), No. 2 (35), July 2012, pp.177-180

39. (2012) Fratu, A.: Simulation of the Robot Systems for Virtual Prototyping in

Assembly Operations. RECENT Journal - ISSN 1582-0246, Vol. 13 (2012), No. 3

(36), November 2012, pp.279-282

40. (2012) Fratu, M., Fratu, A.: Intelligent Design Solutions for Intelligent Buildings.

RECENT Journal - ISSN 1582-0246, Vol. 13 (2012), No. 3 (36), November 2012,

pp.283-286

41. (2013) Fratu, A. : Robot collision avoidance using the programming through

imitation, Revue RECENT, ISSN 1582 – 0246, Vol. 14 (2013), No. 2 (38), July 2013,

pp.87-92.

42. (2013) Fratu, A., Fratu, M.: Physical Robots Programming by Imitation Using

Virtual Robot Prototypes, Bulletin of the Transilvania University of Braşov, Series I:

Engineering Sciences • Vol. 6 (55) No. 2 – 2013, pp. 51-58.

43. (2013) Fratu, A., Dambrine, M., Vermeiren, L., Dequidt, A.: An analysis of the

reciprocal collision avoidance of cooperative robots, In: Journal Automation, Control

and Intelligent Systems, 2013; 1(3): 75-84, Published online July 30, 2013, Vol. 1,

No. 3, 2013, pp. 75-84. doi: 10.11648/j.acis.20130103.17

44. (2014) Fratu, M., Fratu, A.: Visual programming technique and virtual prototyping

technology used in civil engineering, RECENT Journal - ISSN 1582-0246, Vol. 15

No. 2 (42), July 2014, pp.97-101.

45. (1996) Fratu,A. , ş.a. - Cuplaj cu sens unic pentru demaroare de autovehicule.

Lucrările celei de-a 8-a Conferinţe cu participare internaţională CONAT‟96, 13 - 15

noiembrie, 1996, Braşov, p. 303 - 308.

46. (1997) Helerea, E. , Fratu, A. , ş.a. - Baterii de acumulatoare auto - tradiţie şi

inovaţie. Lucrările celei de-a 9-a Conferinţe cu participare internaţională CAR‟97,

noiembrie, 1997, Piteşti. P 13-18.

47. (1998) Fratu, A. - The phase - modulation in control of the electrical corrugater.

Proceedings of the 6-th International Conference OPTIM‟98, 14 - 15 may, 1998,

Braşov, p. 439 - 442.

48. (2000) Ilie, A., Laura Rab, L., Fratu, A., Ilie, R. : ”Electronic simulator for a

frequencially generated reference angular error”. The 7-th International Conference

OPTIM‟2000, may 11-12, Braşov, Romania.

49. (2002) Rab Laura, Fratu A. s.a. : Temperature Measurement Methods Based on

Resistor Thermal Noise Voltage, Proceedings of the 8th

International Conference on

Optimization of Electrical and Electronic Equipments, Brasov, May 16-17, 2002,

p.717-720.

50. (2002) Fratu, A., s.a : Elements de definition du moteur pour servosystemes,The VI-

th International Conference on Precision Mechanic and Mechatronics, COMEFIN-6,

Brasov, 2002, October 10-12, p.79-84.

51. (2006) Fratu, A.: Progress and Trends in the University-Industry Co-operation for

foreign Students- case study. Proceedings ‟‟European Models of Synergy between

Page 107: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

107

Teaching and Research in Higher Education‟‟, Technical University of Tallinn, 4-6

may 2006, p.81-84.

52. (2007) Fratu M. et Fratu, A.: Increasing of the trusting level in the measuring

systems using the redundant information. The International Conference on Library

Science and Science of Information, Brasov 2007, ISBN 978-973-635-973-6, p.169-

172.

53. (2007) Fratu, A.et Fratu M.: Automatically building of the virtual images . The

International Conference on Library Science and Science of Information, Brasov

2007, ISBN 978-973-635-973-6, p.163-168.

54. (2008) Fratu,A. and Fratu, M. : Some considerations regarding the simulation of the

robots movement in the environment with constraints. The International Conference

on Computing and Solutions in Manufacturing Engineering (CoSME), Brasov,

Romania. Academic Journal of Manufacturing Engineering, ISSN 1583-7904, ISSUE

1/2008, p.148-153. www.eng.upt.ro/auif/

55. (2009) Fratu, A., Fratu M., Brethe, J.F., Cachard, O. : Modélisation et animation de

systèmes robotises dans l'environnement Delphi, Colloque francophone ‟‟Evaluation des

Performances et Maîtrise des Risques Technologiques pour les systèmes industriels et

énergétiques”, 28 – 29 mai 2009, p. 105-112, Université du Havre – France.

56. (2009) DEQUIDT A., VERMEIREN L., FRATU A.: Design of flexible drive systems

for feed forward control: a mechatronic approach. 18th International Workshop on

Robotics in Alpe-Adria-Danube Region, ISBN 978 -606-521-315-9, Brasov, Romania,

mai 25-27, 2009.

57. (2009) FRATU A., SISAK F., VERMEIREN L., DEQUIDT A.: Modelling and robot

simulation in Delphi environment. 18th International Workshop on Robotics in Alpe-

Adria-Danube Region, ISBN 978 -606-521-315-9, Brasov, Romania, mai 25-27, 2009.

58. (2009) FRATU, A., DEQUIDT, A., VERMEIREN, L.: Haptic Devices for Desktop

Virtual Prototyping Applications . The 3rd International Conference “Computational

Mechanics and Virtual Engineering” - COMEC 2009, p. 248-253.

59. (2009) FRATU, A., VERMEIREN, L., BRETHE, J-F. : Study of the assembly motion

for a virtual assembly system – International Conference on Economic Engineering

and Manufacturing Systems (ICEEMS) , 26-27 November 2009, Brasov, Romania

60. (2010) Fratu, A., Vermeiren, L., Dequidt, A.,: Using the Redundant Inverse

Kinematics System for Collision Avoidance. The 3rd

International Symposium on

Electrical and Electronics Engineering – ISEEE-2010, 16-18 sept. Galati, Romania,

Proceedings ISBN 978-1-4244-8407-2, pp.88-93.

61. (2011) Fratu, A., Dequit, A., Vermeiren, L.: Virtual Scissors for Haptic Portable

Torque-Feedback Device, Proceedings of the Remote Engineering and Virtual

instrumentation Conference - REV 2011, pp.10-15.

62. (2011) Dang, Q. V., Dequidt, A., Vermeiren, L., Fratu, A., Dambrine, M. : Stability

analysis of haptic interface: effects of dynamic parameters. Proceedings of the Remote

Engineering and Virtual instrumentation Conference - REV 2011, pp. 137-143.

63. (2011) Fratu, A.: A Multi-Robot System for Assembly Tasks in Automotive Industry.

Proceedings of the 4th International Conference Computational Mechanics and

Virtual Engineering COMEC 2011, pp.229-234.

64. (2011) Fratu, A.: Simulation of the Maintenance Analysis for Virtual Steering

Training in Automotive Industry. Proceedings of The 4th International

ConferenceComputational Mechanics and Virtual Engineering COMEC 2011”, pp.

235-238.

65. (2011) Fratu, A.: Robotic Manufacturing Systems – Modelling and Simulation,

Proceedings of the International Conference on Economic Engineering and

Page 108: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

108

Manufacturing Systems (ICEEMS), 26-27 Nov. 2011, RECENT, ISSN 1582 – 0246,

pp.287-290.

66. (2011) Fratu, A.: Actuators as Mechatronics Objects for Motion Software Control.

Proceedings of the International Conference on Economic Engineering and

Manufacturing Systems (ICEEMS), 26-27 Nov. 2011, RECENT, ISSN 1582 – 0246,

pp.291-294.

67. (2011) Fratu M. and Fratu, A.: A Multi-Robot System for Assembly Tasks in Wood

Products Manufacturing Industry. Proceedings of the International Conference on

Economic Engineering and Manufacturing Systems (ICEEMS), 26-27 Nov. 2011,

RECENT, ISSN 1582 – 0246, pp.295-298.

68. (2012) Fratu, A., Ilea, D.: An Analysis of the Reciprocal Robots Collision Avoidance

Proceeding of the 13th International Conference on Optimization of Electrical and

Electronic Equipment (OPTIM), Digital Object Identifier:

10.1109/OPTIM.2012.6231995, IEEE CONF. PUBLICATIONS, pp.1511-1517.

69. (2013) Fratu, A,, Dambrine, M,: An analysis of the robot collision avoidance using

the programming through imitation, The International Conference on Manufacturing

Systems - ICMaS 2013, 14-15 November, 2013, Bucharest, Romania. In: Proceedings

in Manufacturing Systems, ISSN 2067-9238, Volume 8, Issue 2, 2013, pp.111-116.

70. (2014) Fratu, A., Fratu, M.: Imitation-Based Motion Programming for Robotic

Manipulators, Proceedings of 14th International Conference on Optimization of

Electrical and Electronic Equipment OPTIM 2014. May 22-24, 2014, Brasov,

Romania, ISBN 978-1-4799-5183-3/14/$31.00 ' 2014 IEEE, pp. 770-775

71. (2014) Fratu, A., Riera, B., and Vrabie, V.: Predictive strategy for robot behavioral

control, The International Conference on Manufacturing Systems - ICMaS 2014, 14-

15 November, 2013, Bucharest, Romania - accepted to be published.

72. (2014) J.P. Becar, M. Fratu, A. Fratu, J.C. Canonne EXAMPLE BASED

LEARNING FOR VIRTUAL PROTOTYPING ENGINEERING, EDULEARN14

Proceedings OF THE 6th International Conference on Education and New Learning

Technologies, ISBN: 978-84-617-0557-3, ISSN: 2340-1117, 7-9 July, 2014,

Barcelona, Spain, pp. 6027-6034

73. (2015) J.P. Becar, J.C. Canonne, E. Cartignies, M. Fratu, A. Fratu, M. Iwanowski,

W. Czajewski

THREE EXPERIMENTS OF LEARNING BY EXAMPLES,

INTED2015 Proceedings 9th International Technology, Education and Development

Conference, ISBN: 978-84-606-5763-7, ISSN: 2340-1079, 2-4 March, 2015, Madrid,

Spain, pp. 1293-1302

74. Barzel, R. et al.: Plausible motion simulation for computer graphics animation. In:

Proceedings of the Eurographics Workshop on Computer Animation and Simulation,

ISBN:3-211-82885-0, New York, December 1996, Springer-Verlag Publisher, p.183–

197.

75. Bruzzone, L.E., Molfino, R.M., Zoppi, M. and Zurlo, G: The PRIDE prototype:

control layout of a parallel robot for assembly tasks, in Proceeding. of the International

Conference Modelling, Identification and Control MIC2003, 2003, pp. 606-611.

76. C. Fulgenzi, A. Spalanzani, C. Laugier, “Dynamic obstacle avoidance in uncertain

environment combining PVOs and occupancy grid”. In Proc. IEEE Int. Conf. on

Robotics and Automation, pp.1610–1616, 2007.

77. C. Spence and A. Gallace: Recent developments in the study of tactile attention.

Canadian Journal of Experimental Psychology, 61(3), pp. 196–207, 2007.

78. C. Spence and C. Ho: Multisensory warning signals for event perception and safe

driving. Theoretical Issues in Ergonomics Science, 9(6): pp. 523–554, 2008.

Page 109: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

109

79. Calinon, S., Guenter, F. and Billard, A.: On learning, representing and generalizing a

task in a humanoid robot, In IEEE Trans. on Systems, Man and Cybernetics, Part B,

vol. 37, no. 2, pp., 286-298, 2007

80. Cetinkunt S. 1991. Optimal design issues in high-speed high-precision motion servo

systems, Mechatronics, Vol. 1 (2), pp. 187-201.

81. Chiacchio, P., Chiaverini, S., Sciavicco, L., Siciliano, B. Global task space

manipulability ellipsoids for multiple arm systems. In: IEEETransactions on Robotics

and Automation, 7 (1991), p.678-685.

82. Choi, S.H, Chan, A.M.M.: A virtual prototyping system for rapid product

development. Computer Aided Design Journal, Vol. 36, Issue 5, April 2004, p. 401-

412, ISBN 1-58113-506-8, Saarbrücken, Germany

83. Corke, P.I.: A robotics toolbox for MATLAB, IEEE Robot Automation Magazine 3(1),

p. 24–32, 1996.

84. D. Hennes, D. Claes, W. Meeussen, K. Tuyls, “Multi-robot collision avoidance with

localization uncertainty”, In: Proceedings of the 11th International Conference on

Autonomous Agents and Multiagent Systems, Conitzer, Winikoff, Padgham, and van

der Hoek (eds.), June, 4– 8, 2012, Valencia, Spain.

85. D. M. Stipanovic, P. F. Hokayem, M. W. Spong, D. D. Siljak, “Cooperative Avoidance

Control for Multiagent Systems”. In: ASME Journal of Dynamic Systems

Measurement and Control, Vol.129, pp. 699–707, 2007.

86. D. Silver. Cooperative path finding. In The 1st Conference on Artificial Intelligence

and Interactive Digital Entertainment (AIIDE‟05), pages 23–28, 2005.

87. Dequidt A., Valdès E. and Castelain J.M. 1999. Nombres adimensionnels pour la

conception intégrée des mécanismes commandés, 14ème Congrès Français de

Mécanique, CD Rom.

88. Ellis G. and Lorenz R. D. 2000. Resonant load control methods for industrial servo

drives, IEEE Industry Applications Society, Annual Meeting, Rome, Italy, October 8-

12.

89. F. Gosselin and A. Riwan : Design of a wearable haptic interface for precise finger

interactions in large virtual environments. Proc. World Haptics Conference, Pisa,

Italy, 18-20 March 2005, ISBN: 0-7695-2310-2.

90. F. Nagata, I. Okabayashi, M. Matsuno, et al., “Fine gain tuning for model-based

robotic servo controllers using genetic algorithms,” Procs. of the 13th Int. Conf. on

Advanced Robotics, pp. 987–992, 2007.

91. F. Nagata, K. Kuribayashi, K. Kiguchi and K. Watanabe, “Simulation of fine gain

tuning using genetic algorithms for model based robotic servo controllers,” Procs. of

the IEEE Int. Symp. on Computational Intelligence in Robotics and Automation,

pp.196–201, 2007.

92. F. Nagata, Y. Kusumoto, Y. Fujimoto and K. Watanabe, “Robotic sanding system for

new designed furniture with free formed surface,” Robotics and Computer-Integrated

Manufacturing, vol. 23, no. 4, pp. 371–379, 2007.

93. Feliu V., Pereira E., Díaz I. And Roncero P. 2006. Feedforward control of multimode

single-link flexible manipulators based on an optimal mechanical design, Robotics

and Autonomous Systems, Vol. 54 (8), pp. 651-666.

94. Fernandez-Madrigal, J-A. et al.: A software engineering approach for the development

of heterogeneous robotic applications. Proceeding of Robotics and Computer-

Integrated Manufacturing,no.24 (2008) p.150–166.

95. Fernando, T., Wimalaratne, P., and Tan, K., 1999, “Constraint-Based Virtual

Environment for Supporting Assembly and Maintainability Tasks,” Proceedings of the

1999 ASME Design Technical Conference and Computers in Engineering Conference,

DETC99/CIE-9043, Las Vegas, Nevada, September 11-15.

Page 110: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

110

96. Fischer, J., Breithaupt, R., Bode, B.: Adaptive and economic data representation in

control architectures of autonomous realworld robots. Artificial Life and Robotics,

Vol. 6, No. 4, December 2002, p. 200-204, ISSN 1433-5298, Springer Japan Publisher

97. Fleury, G., Lacomme, Ph., Tanguy, A.: Simulation a eveniments discrets. Publisher:

EYROLLES, 2006.

98. Fratu A., Renaux, D., Fratu Mariana: Simulation and Control of the Parallel Robots‟

Dynamics. Bulletin of the TRANSILVANIA University of Brasov, ISSN 2065-2119,

VOL.1 (50), Series I, p.343-350 Caccavale, F., Chiacchio, P., Chiaverini, S. : Task-

space regulation of cooperative manipulators. In: Automatica 36 (2000), p.879-887.

99. Fratu A., Renaux, D., Fratu Mariana: Simulation and Control of the Parallel Robots‟

Dynamics. Bulletin of the TRANSILVANIA University of Brasov, ISSN 2065-2119,

VOL.1 (50), Series I, p.343-350 Dun, X. et al.: Design of a flexible robot assembly

demo system. Proceeding of the 4th

World Congress on Intelligent Control and

Automation, 2002, vol. 2, p. 1343-1346, ISBN 0-7803-7268-9

100. Fratu, A. and Fratu M.: Visual Programming in the Delphi Environment with

Applications in Robotics. Transilvania University Press, Brasov, ISBN 978-973-598-

315-4, 2008 si 2011.

101. Fratu, A. et al: Industrial Robots in a Flexible Assembly System. RECENT Journal,

Vol.10, No.1 (25), p. 28-31,ISSN 1582-0246, Brasov, Romania

102. Fratu, A., Fratu M.: Programarea vizuală în mediul Delphi - cu aplicaţii în Robotică.

Editura Universitatii Transilvania Brasov, 2008.

103. Fratu, A., Fratu, M.: Visual programming in Delphi medium with Applications in

Robotics (Programarea vizuală în mediul Delphi cu AplicaŃii în Robotică)

“Transilvania” University Press, ISBN 978-973-598-315- 4, Brasov, Romania, 2008

(in Romanian)

104. Fratu, A., Fratu, Mariana: The problem of operational space control for redundant

robots. RECENT, Vol. 6, No. 3(15), December 2005, p. 52-55, ISSN 1582 – 0246,

Brasov, Romania

105. G. Schreiber, Ch. Ott, G. Hirzinger: Interactive Redundant Robotics: Control of the

Inverted Pendulum with Nullspace Motion, Proceedings of the 2001 IEEE/RSJ

International Conference on Intelligent Robots and Systems Maui, Hawaii, USA, Oct.

29 – Nov. 03, 2001

106. Galoppo, N.: Animation, Simulation, and Control of Soft Characters using Layered

Representations and Sampled Physics-based Methods, Dissertation submitted to the

faculty of the University of North Carolina Chapel Hill, 2008.

107. Gillespie,R.B. et al: Haptic Interface for Hands-On Instruction in System Dynamics

and Embedded Control, In 11th Symposium on Haptic Interfaces for Virtual

Environment and Teleoperator Systems, 2003, pp. 410-415.

108. Good M.C., Sweet L.M. and Strobel K.L. 1984. Dynamic models for control system

design of integrated robot and drive systems, ASME J. of Dyn. Syst., Meas., and Cont.,

Vol. 107, pp. 53-59.

109. Gowda, S., Jayaram, S., and Jayaram, U., 1999, “Architectures for Internet-based

Collaborative Virtual Prototyping,” Proceedings of the 1999 ASME Design Technical

Conference and Computers in Engineering Conference, DETC99/CIE-9040, Las

Vegas, Nevada, September 11-15.

110. Gregory, A. et al.: A Framework for Fast and Accurate Collision Detection for

HapticInteraction. In Proceedings of the IEEE Virtual Reality (VR‟99), Pg. 38-45,

1999.

111. Grimes, D., Chalodhorn, R. and Rao, R.: Dynamic imitation in a humanoid robot

through non-parametric probabilistic inference, In Proc. Robotics: Science and

Systems (RSS), Philadelphia, Pennsylvania, USA, August 2006.

Page 111: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

111

112. Grimes, D., Chalodhorn, R. and Rao, R.: Dynamic imitation in a humanoid robot

through non-parametric probabilistic inference, In Proc. Robotics: Science and

Systems (RSS), Philadelphia, Pennsylvania, USA, August 2006.

113. Hashimoto, H.: Intelligent interactive spaces - integration of IT and robotics, In:

Proceedings of 2005 IEEE Workshop on Advanced Robotics and its Social Impacts, p.

85–90

114. Heejin, L., Hyunseok, S., Euntai, K., Seungwoo K., Mignon, P.: Variable structure

control of manipulator using linear timevarying sliding surfaces. Proceeding of

International Conferences on Intelligent Robots and Systems, Vol. 2, p. 806-811,

ISBN 0-7803-4465-0, October 1998, Victoria, BC, Canada

115. Helmer, P.: 3D force-feedback wrist, MS thesis, Swiss Federal Institute of

Technology, Lausanne, Switzerland, 2000

116. Hoffmann, H., Pastor, P., Park, D.-H., and Schaal, S.: Biologically-inspired dynamical

systems for movement generation- Automatic real-time goal adaptation and obstacle

avoidance. In International Conference on Robotics and Automation, pp. 2587–2592,

2009, Piscataway, NJ: IEEE.

117. Hosoda, K., Igarashi, K., Asada, M.: Adaptive hybrid control for visual servoing and

force servoing in an unknown environment. IEEE Robotics and Automation

Magazine, Vol. 5, No. 4, p. 39–43, December 1998, ISSN 1070.99321, University of

South Florida, USA

118. Hu, X. and Zeigler, B.P.: Model Continuity to Support Software Development for

Distributed Robotic Systems: a Team Formation Example. Journal of Intelligent &

Robotic Systems, Theory & Application, Vol. 39, Issue 1, January 2004, p. 71-87,

ISSN 0921-0296, Springer Netherlands

119. I. Díaz and J.J. Gil, “Influence of Internal Vibration Modes on the Stability of Haptic

Rendering”, Robotics and Automation, 2008.ICRA 2008. IEEE International

Conference; “Influence of Vibration Modes and Human Operator on the Stability of

Haptic Rendering”, IEEE TRANSACTIONS ON ROBOTICS, VOL. 26, NO. 1,

FEBRUARY 2010.

120. Ijspeert, J., Nakanishi, J. and Schaal, S.: Movement imitation with nonlinear

dynamical systems in humanoid robots. In Proceedings of the IEEE International

Conference on Robotics and Automation, pp.1398-1403, 2002.

121. J. E. Colgate and G. Schenkel, “Passivity of a class of sampled-data systems:

Application to haptic interfaces,” Journal on Robotic System, vol.14, no.1, pp. 37–47,

1997.

122. J. J. Gil, E. Sanchez, T. Hulin, C. Preusche, and G. Hirzinger, “Stability boundary for

haptic rendering: Influence of damping and delay,” in IEEE Int. Conf. Robot. Autom.,

Roma, Italy, April 10-14 2007, pp. 124–129.

123. J. Pettre, J. Ondrej, A.-H. Olivier, A. Cretual, and S. Donikian. Experiment based

modeling, simulation and validation of interactions between virtual walkers. In

Symposium on Computer Animation. ACM, 2009.

124. J. Snape, J. van den Berg, S. J. Guy, and D. Manocha, “Independent navigation of

multiple mobile robots with hybrid reciprocal velocity obstacles”, IEEE/RSJ Int.

Conf. Intelligent Robots and Systems, St. Louis, Mo., 2009;

125. J. Snape, J.van den Berg, S.J. Guy, D. Manocha, J. Snape, S. J. Guy, J. van den Berg,

S. Curtis, S. Patil, M. Lin, and D. Manocha, “Independent navigation of multiple

robots and virtual agents”. In Proc. of the 9th Int. Conf. on Autonomous Agents and

Multi agents Systems (AAMAS 2010), Toronto, Canada, May 2010.

126. J. van den Berg, D. Wilkie, S. Guy, M. Niethammer, D. Manocha, “LQG-Obstacles:

Feedback control with collision avoidance for mobile robots with motion and sensing

Page 112: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

112

un-certainty.” IEEE Int. Conf. on Robotics and Automation, River Centre, Saint Paul,

Minnesota, USA, May 14-18, 2012, pp. 346- 353.

127. J. van den Berg, S. Guy, M. Lin, D. Manocha, “ Reciprocal n-body collision

avoidance”. In: Proc. Int. Symposium on Robotics Research, 2009

128. J. van den Berg, S. J. Guy, M. Lin, and D. Manocha, C. Pradalier, R. Siegwart, and G.

Hirzinger, ”Reciprocal n-body collision avoidance”, Robotics Research, The 14th

International Symposium ISRR, Springer Tracts in Advanced Robotics, vol. 70,

Springer-Verlag, May 2011, pp. 3-19.

129. J.E. Colgate and J.M. Brown, “Factors Affecting the Z-Width of a Haptic Display”,

Proceedings of the IEEE 1994 International Conference on Robotics & Automation,

pp. 3205-10, San Diego, CA, May 1994

130. J.J Gil, M.J. Puerto, I. Díaz and E.Sánchez, “On the Z-Width Limitation due to the

Vibration Modes of Haptic Interfaces”, G. Hirzinger, A. Albu-Schaffer, M. Hahnle, I.

Schaefer, and N. Sporer: On a new generation of torque controlled light-weight

robots. IEEE International Conference of Robotics and Automation, pp. 3356-3363,

2001.

131. J.J. Gil, A. Avello, Á. Rubio, and J. Flórez, “Stability analysis of a 1 DOF haptic

interface using the Routh–Hurwitz criterion”, IEEE transactions on Control Systems

Technology, vol. 12, no. 4, july 2004, pp. 538-588.

132. K. Gold. An information pipeline model of human-robot interaction. In Proceedings of

the 4th ACM/IEEE international conference on Human robot interaction, pp. 85- 92,

New York, NY, USA, 2009. ACM. ISBN 978-1-60558-404-1.

doi:http://doi.acm.org/10.1145/1514095.

133. K. O. Arras, J. Persson, N. Tomatis, R. Siegwart, “ Real-Time Obstacle Avoidance for

Polygonal Robots With a Reduced Dynamic Window” in Proc. IEEE Int. Conf. on

Robotics and Automation, Washington DC, May 2002, pp.3050-3055.

134. Khalil W. and Liegeois A. 1984. Génération des mouvements optimaux en robotique,

RAIRO Automatique/Systèmes analysis and Control, Vol. 18(1) pp. 24-39.

135. Khalil, W., Ibrahim, O.: General Solution for the Dynamic Modeling of Parallel

Robots, In Journal of Intelligent and Robotic Systems 49 (2007) Nr.1, p.19-37.

136. Kim J, Kim K, Choi K, Lee J.Y.: Solving 3D geometric constraints for assembly

modeling. International Journal Advanced Manufacturer Technologies, Vol. 16, No 11,

September 2000, p. 843-849, ISSN 0268-3768, Springer Netherlands.

137. Kober, J., Peters, J. : Learning motor primitives for robotics. In Proceedings of the

IEEE International Conference on Robotics and Automation, pp. 2112–2118, 2009,

Piscataway, NJ: IEEE.

138. Kulvicius, T., Ning, K., Tamosiunaite, M., and Worgotter, F. : Joining movement

sequences- Modified dynamic movement primitives for robotics applications

exemplified on handwriting, In IEEE Transactions on Robotics, 28(1), pp. 145–

157, 2012.

139. L.D., Magillo, P., Puppo, E., Sobrero, D.: A multiresolution topological representation

for non-manifold meshes. Computer Aided Design Journal, Vol. 36, Issue 2,June,

2004, p. 141–59, ISBN 1-58113-506-8,Saarbrücken, Germany

140. Lambrechts P., Boerlage M. and Steinbuch M. 2005. Trajectory planning and

feedforward design for electromechanical motion systems, Control Engineering

Practice, Vol. 13 (2), pp. 145-157.

141. Lawrence, D. A. et al: Human Perceptual Thresholds of Friction in Haptic Interfaces,

In Proceedings of the ASME Dynamic Systems and Control Division, DSC-Vol.64,

Pp. 287-294, ASME 1998

Page 113: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

113

142. Li, Y.T., Hu, S.M., Sun, J.G.: A constructive approach to solving 3-D geometric

constraint systems using dependence analysis. Computer Aided Design Journal, p.

Vol. 34, Issue 2, February 2002, 97-108, ISSN 0010-4485,Elsevier

143. Lippiello, V. et al.: Robot Force/Position Control with Force and Visual Feedback.

Proceedings of the European Control Conference 2007, ISBN: 978-960-89028-5-5,

Kos, Greece, July 2-5, 2007

144. Luo_and S. Ahmad: Mathematical Analysis of the Joint Motion of Redundant

Robots Under Pseudo-Inverse Control. ECE Technical Reports Purdue Libraries

TR-EE 92-12, March 1992, Purdue University West Lafayette, IN 47907, USA

145. M. Turpin, N. Michael, V. Kumar,” Trajectory planning and assignment in multi robot

systems”. Proc. Workshop on Algorithmic Foundations of Robotics, 2012.

146. McNeely, W. A. et al: Six degree-of-freedom haptic rendering using voxel sampling,

In Proceedings of the 26th annual conference on Computer graphics and interactive

techniques, p.401-408, July 1999.

147. Merlet, J.-P.: Parallel Robots (Second Edition). Springer Publisher, Dordrecht, The

Netherlands, 2006.

148. Michel, O.: Professional Mobile Robot Simulation. International Journal of Advanced

Robotic Systems, vol. 1, no. 1, p. 39-42, 2004

149. Miller, K.: Optimal design and modeling of spatial parallel manipulators. In: Int. J. of

Robotics Research, 23(2004), p.127–140.

150. Mohsen Mahvash, M. et al.: Modeling the Forces of Cutting with Scissors, IEEE

Trans Biomed Eng. 2008 March. ISSN: 0018-9294, Vol.55(3), p. 848 – 856

151. N. Diolaiti, G. Niemeyer, F. Barbagli, and J. K.Salisbury, “Stability of haptic

rendering: Discretization, quantization, time-delay and coulomb effects,” IEEE Trans.

Robot., vol. 22, no. 2, pp. 256–268, 2006

152. N. Galoppo, “Animation, Simulation, and Control of Soft Characters using Layered

Representations and Simplified Physics-based Methods” Dissertation submitted to the

faculty of the University of North Carolina Chapel Hill, 2008.

153. N. K. Govindaraju, S. Redon, M. C. Lin and D. Manocha, “CULLIDE: Interactive

Collision Detection Between Complex Models in Large Environments using Graphics

Hardware”, M. Doggett, W. Heidrich, W. Mark, A. Schilling (Editors), Graphics

Hardware, 2003.

154. Nicolescu, M. and Mataric, M.: Natural methods for robot task learning- Instructive

demonstrations, generalization and practice, In Proc. Intl Joint Conf. on Autonomous

Agents and Multi-agent Systems (AAMAS), Melbourne, Australia, pp. 241-248, 2003.

155. Okamura, A. M. et al.: The Haptic Scissors: Cutting in Virtual Environments.

Proceedings of the 2003 IEEE International Conference on Robotics & Automation

Taipei, Taiwan, September 14-19, 2003; Feeling is believing: Using a Force-

Feedback Joystick to Teach Dynamic Systems, In ASEE Journal of Engineering

Education, Vol. 91, No. 3, 2002, pp. 345-349.

156. P. Chiacchio, S. Chiaverini, L. Sciavicco, B. Siciliano: Closed-loop inverse kinematics

schemes for constrained redundant manipulators with task space augmentation and

task priority strategy, Int. J. Robot. Res. 10(4), 410–425 (1991)

157. Pardowitz, M., Zoellner, R., Knoop, S. and Dillmann, R.: Incremental learning of

tasks from user demonstrations, past experiences and vocal comments, In IEEE Trans.

on Systems, Man and Cybernetics, Part B, vol. 37, no. 2, pp. 322-332, 2007.

158. Parker, L.E.: Distributed Algorithms for Multi-Robot Observation of Multiple Moving

Targets. Journal Autonomous Robots, Vol. 2, No. 3, May 2002, p. 231-255, ISSN

0929-5593, Springer Netherlands Publisher

Page 114: Universitatea Transilvania din Braşov...Abordarea actuală sistematică a proiectării controlerelor pentru sisteme neliniare este metoda liniarizării folosind feedback-ul. Ideea

Habilitation thesis Aurel Fratu

114

159. Pasch K.A. and Seering W.P. 1984. On the drive systems for high-performance

machines, ASME J. of Mechanisms, Transmissions and Automation in Design, Vol.

106, pp. 102-108.

160. Pastor, P., Hoffmann, H., Asfour, T., and Schaal, S.: Learning and generalization of

motor skills, by learning from demonstration. In International Conference on Robotics

and Automation, pp. 763–768, 2009, Piscataway, NJ: IEEE.

161. Piazzi, A. And Visioli, A. 2000. Minimum-time system inversion based motion

planning for residual vibration reduction, IEEE Transactions on Mechatronics, Vol.

5(1), pp. 12–22.

162. Pollard, N., Hodgins, J., Riley, M. and Atkeson, C. : Adapting human motion for

the control of a humanoid robot, In IEEE International Conference on Robotics and

Automation, 2002.

163. Popescu, V. G., Burdea, G. C.: Dextrous Haptic Interface for Jack. In Proceedings of

the ASME Dynamic Systems and Control Division, DSC-Vol. 64, Pp. 189-194,

ASME

164. Q.J. Ge : An inverse design algorithm for a G2 interpolating spline motion, In

Advances in Robot Kinematics and Computational Geometry. Kluwer Academic

Publishers, Norwell, MA, 1994, p. 81- 90.

165. Q.J. Ge : An inverse design algorithm for a G2 interpolating spline motion, In

Advances in Robot Kinematics and Computational Geometry. Kluwer Academic

Publishers, Norwell, MA, 1994, p. 81- 90.


Recommended